The CS

Assistant Professor
Handbook

Vijay Chidambaram

3
4
 To my wife Prashipa,
My children Kumaran and Keerthi,
My parents and family

5
Vijay Chidambaram
The CS Assistant Professor Handbook

© 2022, Vijay Chidambaram

Self-published

All rights reserved.

No part of this publication may be reproduced, stored in a retrieval system, stored in a database
and / or published in any form or by any means, electronic, mechanical, photocopying, recording
or otherwise, without the prior written permission of the publisher.

6
Preface
Welcome to the CS Assistant Professor’s Handbook. In this book, I will attempt to demystify the
job of the tenure-track assistant professor. I will talk about what the job entails, briefly about
how to get such a job, the financial aspects of the job, and then about each big aspect of the
job: students, research, teaching, and service. I will talk about how to do effective outreach, and
finally talk about how to take care of yourselves while on the tenure track. I hope you find this
book useful!

About me
I’m Vijay Chidambaram, an associate professor at the Department of Computer Science at the
University of Texas at Austin. I was born and brought up in Chennai, India. I did my
undergraduate education at the College of Engineering, Guindy (CEG) in Chennai, and then did
my graduate studies at the University of Wisconsin-Madison with Remzi Arpaci-Dusseau and
Andrea Arpaci-Dusseau. I got my Masters degree in 2011, and my PhD in 2015.

At the end of my PhD, I was lucky enough to get a tenure-track assistant professor position at
the Computer Science department at the University of Texas at Austin. I deferred the offer for a
year, doing a post-doc at VMware Research. I then joined UT Austin in Fall 2016, and was
promoted to Associate professor starting Fall 2022.

Why write this book?

I’m writing this book for multiple reasons. First, I’m writing it as a memoir of my own
assistant-professor days, as a sort of documentation of the journey. I was inspired in this regard
by Arun Kumar’s blog post series, “The Secret Lives of Millennial CS Assistant Professors”
(incidentally, Arun and I were room-mates for a while when we were getting our PhDs in
Wisconsin!). Second, I feel like there is no central place where a graduate student who is
interested in an academic job could get this information. There are definitely excellent blog
posts and tweet threads on various aspects of the job (which I’ll link to in this book), but the
information is scattered and not organized. This is one way of giving back to the community,
collecting together all the information and advice I’ve been given over the years. Finally, I feel
like this is information I’ll give my own students and postdocs hopefully multiple times in the
coming years. Following Matt Might’s advice, I’m writing it down since it will be useful for a
wider audience.

10
Who is this book for?
The main audience for this book is folks who just started (or expect to start soon) as an
tenure-track assistant professor in computer science. Reading this the summer before starting
the position, or the first semester on the job, should help you navigate different aspects of this
job. The book should also be useful to senior graduate students who will go on the job market in
one or two years.

The book should also serve as an introduction to the academic life for all students, both
graduate and undergraduate, who are wondering what it is like to be a tenure-track assistant
professor.

Important Caveats
There are a number of caveats I have to mention.

First, this book represents my opinions on the topics covered here. There is a large variety of
opinions on all these topics, and if you ask N professors you will get N answers. There is no
correct answer, and this is just one take on these topics.

Second, this book is based on my own experiences on the tenure track, and as such it is going
to be based on what it is like to be a tenure-track assistant professor in computer science at an
R1 public research university in the USA. I’d imagine experiences vary widely at an
undergrad-focused university, or in Europe, or in a different discipline like biology. Even among
R1 public universities in the USA, there are significant differences. So please take that into
account when reading the book.

Finally, I’m not claiming to be the expert on any of the topics covered here. It only represents my
own experiences and opinions. It is quite likely there are better ways of doing many of the things
I discuss in the book. I’ll link to expert advice and books wherever I can.

Scope of the book

Folks can sign for book updates using this Google form, and I’ve been getting requests for a lot
of things where I don’t really have any expertise (such as startups). For version 1.0 of this book,
my plan is to just talk about my own experiences, but I do plan to later supplant this with
interviews or guest chapters from other folks.

11
Chapter 1: The Assistant
Prof Job
Welcome to The CS Assistant Professor Handbook. My assumption in this book is that you, dear
reader, are an undergraduate or graduate student, who is interested in academia and wants to
know more about the assistant professor job. If you are a postdoc or someone who just began
an assistant professor position, you might find I’m presenting information that is rather
well-known to you. Hold on, and later chapters should be of interest!

So let us start with the basics – what is a tenure-track assistant professor position?

We can broadly classify positions in academia into two kinds. Positions with indefinite tenure, or
tenured positions in short, are those where the holder of the position cannot be fired unless
there are extraordinary circumstances such as fraud or abuse. It is essentially a job for life, with
a stable paycheck. Positions without tenure do not have this protection, in that the holder can
be fired for causes such as bad economic conditions. For example, consider the Adjunct
position, where the holder is typically on a multi-year contract that has to be renewed
periodically. Teaching instructors also typically do not have tenure.

Why does tenure exist? There are a number of reasons. First, it is meant to encourage
professors to work on multi-year high-risk, high-reward projects. Some research requires
investing multiple years, and working on hard problems always comes with the risk it doesn’t
work out. If the professors are evaluated on a short time frame, they would naturally not work on
such projects. Thus, tenure is meant to encourage such exploration.

Second, it is meant to protect academic freedom, to prevent the university management from
firing a professor who has different political or academic views. As such, a tenured professor
can publicly criticize the management without fear of immediate dismissal. This is a rare
privilege, not accorded to many professions. For example, if a software engineer publicly
criticizes his CEO, they can expect to get fired (especially under “at will” employment
arrangements that are common in the USA).

Finally, it is meant to be a perk to entice professors to academia. In fields like engineering or

computer science, a professor gets paid anywhere between 5 percent to 50 percent what their
peers make in industry; the stability of tenure is a perk that cannot be found in industry. Even
though most CS professionals can always find another job, nobody likes working with a sword
hanging over their neck, especially in times of economic turmoil. For example, Microsoft in 2014
decided to suddenly shut down their highly regarded Silicon Valley research lab, firing everyone
12
except a handful of folks. Although everyone involved found new jobs quickly1, there is no
denying this was a stressful experience.

There is ongoing debate whether academic positions should have tenure. Opponents suggest
that tenure leads to complacency, that tenure makes it harder to fire professors who engage in
abusive behavior, and that since computer science is a lucrative field, that tenure does not have
the same draw as in other sciences. Proponents say that it is vital for academic freedom, and
that lack of tenure will stifle critical voices.

The academic ladder. For the tenure-track, the entry-level position is the Assistant Professor
position. Sometimes, PhD holders fresh out of their grad school are hired as Assistant
Professors; others work in industry for sometime before applying for the Assistant Professor
position. In some fields such as theory, a postdoc might be expected before being hired into the
Assistant Professor position.

Assistant Professors are evaluated after a certain number of years to see if they should be given
tenure. The number of years varies from university to university: it is six years at UT Austin, but
ten years at CMU. Sometimes Assistant Professors get promoted to Associate Professors with
tenure; at other places, the promotion to Associate comes first, and later they are evaluated for
tenure.

Associate professors are typically promoted to full professors after 6-8 years (again, this varies
from place to place). Full professorships always come with tenure.

1.1 What does the job consist of?

The assistant professor job varies a lot depending upon the university where it is performed.
There is a wide spectrum of universities, with R1 research institutions on one end, and
teaching-focussed 4-year colleges on the other end. Your experience as an assistant professor
at these two endpoints would be quite different, in terms of what is expected in research and
teaching, the resources offered, and the time given to achieve expectations. For example, at an
R1 institution, you might be teaching one course a semester. At a teaching institution, you might
be expected to teach three. At some other places, you are expected to do research on top of
teaching multiple courses!

In this book, I’m primarily going to be talking about the job at an R1 institution. At an R1
institution, the job consists of at least the following five parts (corresponding to each of the
chapters in this book):

1) Research. You are expected to perform independent research and publish in your
community’s accepted venues. Each department will have different expectations
regarding where it is acceptable to publish, and how much you are expected to publish.

1
Incidentally, a bunch of folks from MSR Silicon Valley got together and started the VMware Research group, where
I did my postdoc and where I am doing my sabbatical
13
 2) Teaching. You are expected to teach at least one course every fall and spring semester.
Some universities require you to teach at least one undergraduate course every year.

3) Students. You are required to recruit, mentor, and train students to be independent
researchers during the course of their PhD.

4) Funding. Closely related to #3 (but not always2), you are expected to raise funds via grant
proposals to support your research group and your research agenda. This is primarily
done through proposals to the National Science Foundation (NSF), the The Defense
Advanced Research Projects Agency (DARPA), and other funding sources such as the
Army. Yet another source of funding, note it is not as large as government sources, is
industry.

5) Service. You are expected to serve on program committees, review journal papers, and in
general, be a “good citizen” of both your research community, and your department. You
can expect to serve on at least one committee each semester at your department – good
departments will put new assistant professors on committees such as PhD admissions.
You are expected to write letters of recommendation for your students (and eventually
your peers).

Thus, the assistant professor job is a multi-faceted one, requiring you to learn and master
several different aspects in a short time frame. It is similar to starting a company and being the
CEO – you have to worry about funding, recruiting folks, getting the word out about your
company etc. You have a lot to worry about, but you also have a lot of freedom. You don’t have a
“boss” per se, and you can determine what kind of research you want to do, and what kind of
group you want to build.

Given how much the job expects of you, you might reasonably determine that it pays a lot. Let
us talk about that next.

1.2 Financial aspects

How much do assistant professors get paid? It varies quite a bit based on the university. In
general, private universities, such as Princeton, pay a lot more than public universities (which
have strict rules regarding how much your salary can be).

You might be surprised to learn this: at R1 universities in the USA, assistant professors are only
paid for nine months of the year. The other three months you have to support yourself on your
grants. Therefore, if you don’t have grants, you don’t get paid for three months of the year
(typically the summer). There are even schools that put restrictions on what kind of funds you
can use for the summer – for example, they might prevent you from using your startup funds
(funds given to kickstart your research program) for summer salary.

2
Some institutions do not care as long as you are able to fund your students. At other universities, you are
expected to bring in a certain dollar amount via grants that goes towards a common pool that is used to fund
research.
14
If you are thinking “This seems outrageous!”, you are not alone. Some professors, including me,
feel that the nine-month salary is simply a way for universities to underpay professors. The only
other job I can think of where you are partially responsible for your salary despite working for an
employer is being a waiter (where you have to work for tips). When the university gives you a
9-month offer, it is saying “we agree one hour of your time is worth X”. But if you can’t find
external funding, the university can get away with paying you X*9/12 = X*3/4 for each hour of
your work. The value of your time should not depend on external funding. Given that grants are
highly competitive, it is quite likely that many professors are unable to cover their summer
months from grants; thus, universities get three months of labor without explicitly paying for it.

Professors on the other side of the argument say that they consider the 9-month salary to be a
12-month salary. Thus, they expect to get paid only X* 3/4 (to continue my prior example) and
feel that any more is a bonus. While it is certainly true that you could live comfortably off a
9-month salary in some parts of the USA, I do not think this would be feasible in high
cost-of-living areas. More than that though, the question, in my opinion, is not whether
professors are paid enough, the question is are professors paid what they deserve given their
expertise and years of experience? I do not think so (see Pros and Cons).

Alright, with that being said, what do assistant professors make? The nine-month salary for
assistant professors in CS typically tends to be in the $90K to $150K range (see the CRA
Taulbee Survey Figure S6). As you would expect, the higher end is private universities. So let us
say you make $120K over nine months, you might get around $160K annually (if you have
summer funding via grants). You will get about 1%-4% raises every year, for cost of living
adjustments.

How does this compare to salaries in industry? If you are a computer science PhD joining
industry in 2022, you can expect to make about $160K in your base salary, with around
$100-200K of stock (typically vesting over 4 years), and perhaps a $10-20K signing bonus. Note
that is the lower end – I have heard of fresh PhDs getting million dollar packages (with stock
vesting over 4 years). You can expect to get around 5-6% increments each year in industry,
depending on your job performance.

So the difference in compensation between industry and academia is quite large, and tends to
get wider and wider over time. About ten years after the PhD, a professor might be earning
about $130-230K in salary, while their industrial peers might be getting 2x to 5x that much in
total annual compensation.

This is not to say that professors cannot become wealthy. For example, Ion Stoica and Scott
Shenker recently donated $25M each to the Berkeley computer science department. In most
cases, such wealth is the result of co-founding a highly successful company. It is becoming
more and more common for computer science professors to start-up companies based on their
research and expertise. Startups aside (which you can also do in industry), typically, professors
earn significantly less than their industry peers.
15
1.3 Consulting
One aspect of the professor job that helps professors earn more money is consulting. When you
consult for a company, you are lending your expertise to them, as a contractor, for a specific job.
You are not an employee of the company (and thus do not receive employee benefits such as
medical insurance or stock options). You get paid an hourly rate that you negotiate with the
company.

At many R1 universities in the USA, you are allowed to consult 1 day/week. Thus, you can
supplement your income with the money from consulting. Consulting rates range from hundreds
to thousands of dollars per hour (depending upon your expertise). Getting started with
consulting is a little tricky, since there needs to be a match between your expertise and what the
company is looking for. But such consulting relationships can be mutually beneficial once
formed.

Unless you have prior connections, it is unlikely you will begin consulting when you start the
assistant professor position. Thus, it is not a part of your income that you should take for
granted. But with time and networking, it is possible to get some extra income from consulting.

Consulting is a bit trickier if you are on a visa, though it is not impossible. It is possible to do
consulting before obtaining permanent residency in the USA. I would recommend talking to a
good immigration lawyer about this.

To talk about my own consulting experience, I have been consulting for VMware since I joined
UT Austin. I had a prior relationship with VMware from my internships and post-doc. I had
started some projects as a post-doc that I continued to work on as a consulting researcher. The
consulting relationship has been super useful, both from a financial point of view, and being able
to learn about problems engineers face within VMware.

One of the key things to remember when doing consulting work is to keep your university
intellectual property (IP) and your consulting job IP separate. University resources cannot be
used for your consulting job, and company resources cannot be used for your university
research. It is best to have different projects that you work on, so that it is clear what hat you are
currently wearing when working on a project.

Can you consult for two different companies at the same time? It really depends on the
companies involved and what you are consulting on. For example, rival companies might not
want to consult with you about the same topic. However, consulting for different companies on
different projects is normal; again, make sure to keep the IP separate!

Fears of IP leaking across your projects might cause companies to not form a consulting
relationship with you, if they know you already have such a relationship with another company.
As a result, while you can theoretically consult with multiple companies, my experience has

16
been consulting with one company at a time, building a long-term relationship with that
company.

Can you consult for company X, and collaborate (where you are not paid) with company Y?
Yes. For example, I consult for VMware, but I have simultaneously collaborated with researchers
in Microsoft and Google. The key is the IP and the role: as a professor, you are free to
collaborate with any company: the IP is shared among the university and the industrial
collaborators. If the collaborative research leads to patents, they must be jointly filed by the
collaborators. In short, any work done using university resources gives the university a right to
the resulting IP. However, when you consult, you are not (and should not be) using any university
resources; you are doing it in your private capacity, and the IP is wholly owned by the company.

Another way collaborations happen is via internships. When students intern at a company, they
are using company resources, and the IP they create belongs to that company. Once that IP has
been patented, you can then collaborate with the team in your role as a professor. Thus, you
might be consulting with company X, but your student interns at company Y and through them
you begin a collaboration with company Y.

1.4 Pros and cons of the job

Now that we have talked about what the job entails, let us chat briefly about the pros and cons
of the job. Note that I am obviously biased, since I chose to go into academia. So take my
opinions with a pinch of salt.

First, lets see the pros:

1. Freedom. You don’t have a boss as an assistant professor. Nobody tells you what to work
on at any given day. You can choose what topics you want to research, which students to
hire, who to collaborate with. You can collaborate with different universities, and different
companies. If you are interested in a new topic and want to work on it, nobody will stop
you. It is hard to find such freedom outside of academia.

2. Students. There is a special joy in mentoring students and seeing them grow from
strength to strength. You enjoy both longer-term relationships with your PhD students,
and shorter term relationships with undergraduate students. If you have heard of the
expression “being with students keeps you young at heart”, it is definitely true. This is
primarily why I became a professor.

3. Flexibility. Being a professor allows you to do so many things. You can write a book. You
can give talks to the general public. You can create online courses. You can start a
company. The sheer range is quite amazing.

4. Time with family. This might seem like a weird pro, but as a professor, I have much more
control over my time than many other jobs. It lets me schedule my work so that I can
spend more with my family. For example, when my first kid was born, my department

17
 gave me teaching time off so that I could spend time with him. I spent six months with
my son – irreplaceable time that is harder to get in many other jobs (this is especially
true in the USA where paternity and maternity leave is minimal in many jobs).

Now for the cons:

1. Responsibility. With great freedom comes great responsibility. One of the perks of having
a boss is that somebody tells you what to do – you don’t have this as an assistant
professor, and the burden of responsibility is quite high. Not only do you have to worry
about your own career, your PhD students are entrusting their careers to you. So this job
carries a lot of responsibility.

2. Stress. The first few years of being an assistant professor are stressful. You are learning
different aspects of the job (for many of these aspects you would not have been trained
formally) while under pressure. The good news is that it gets better after the first few
years. This book is also an attempt to reduce stress by better preparing you for the job.

3. Poor financial compensation. If an assistant professor went into industry, they would
easily make 50% to 2x more initially. Over time, being in academia means you are giving
up sizable amounts of money. They say that the financial cost of a PhD is a house (about
$500K). Well, I would say the financial cost of being a professor is a neighborhood
(several million USD). While you will make enough to have a decent living, the opportunity
cost is quite high. In some high-cost-of-living areas, you may not make enough as a
professor to offset the high cost of daycare (especially true in the USA).

4. Funding. If every professor could magically get funds for their research without having to
write proposals, I’m sure everyone could take that option. Pending that, we have to spend
a lot of time and effort on making sure our research group is adequately funded. Adding
to the problem is the fact that NSF funds only about 26% of the proposals it receives. So
you might have to try multiple times to get a proposal funded.

A final con is something more general to academia in the USA as whole, and not specific to the
assistant professor position: a culture of relying on volunteering or unpaid labor. As far I can
see, US academia does not pay anybody (except perhaps the administrators) what their time
and work is actually worth (instead, you get “prestige” or class credit or a line in your CV). This
goes both for you, and your students. You will have to do a lot of unpaid labor in the name of
service; your students will also be expected to volunteer and do unpaid work (not just things like
reviewing, which are necessary in any research community, but things related to running a
conference that should really be handled by professionals). I really wish this was different - this
might change perhaps with time, but right now, this is an aspect you want to consider. Industry
is much better at compensating folks fairly for their time.

Overall, I would say being an assistant professor is something you should choose after careful
consideration. If you have the skills to be an assistant professor, you likely have the skills to be
an entrepreneur or a business founder/owner – these typically pay more for the effort put in.
Make sure that the pros outweigh the cons for you personally so that your motivation is strong.

18
1.5 Why I chose this job
This is a bit of a long story, but I figured I’d share:

I originally joined Wisconsin for a Masters degree. Remzi and Andrea Arpaci-Dusseau hired me
as a graduate research assistant, despite no notable research experience. So I started doing
research with Remzi and Andrea.

At that point, I was not looking to do a PhD or go into academia. My vision for life was pretty
short-term: I wanted to get a high-paying job, buy a car, and enjoy life with my friends.

The research I worked on with Remzi and Andrea got published, and I traveled to Hong Kong to
present it. I was super excited about my first conference talk! After weeks of preparation and
thousands of miles of travel, there were six people in the audience for my talk (it was a
multi-track conference). Of the six, two more were also presenting. At that point, I didn’t really
see the point in academia: I worked so hard for this?

So near the end of my Masters degree, I got a job at Microsoft and I told Remzi and Andrea.
They asked me to consider staying six more months to try to see if a PhD was for me. It was
actually my parents who encouraged me to pursue a PhD – I was actually extremely lucky to get
Remzi and Andrea as advisors, but I was too dumb and un-informed at the time to know what a
big deal this was. So I was lucky twice – to get amazing advisors, and to get amazing parents
who encouraged me to pursue research rather than making money.

Anyway, I stayed for six months, and everything changed. My first paper got published at FAST,
and I gave an amazing experience giving the talk to 500 people in the storage community. That
led to the next paper, and that led to the next, and that’s how I got my PhD.

Until the last two years of my PhD, I had no intention of becoming a prof. But over time, I noticed
that Remzi and Andrea had so much fun at their job, and that a lot of the fun came from
interacting with students. I mentored a few students myself, and really enjoyed the experience. I
taught a class at Wisconsin, and found that I enjoyed doing that as well.

So when I graduated, I figured I would give the academic route a shot. I honestly did not expect
any interview calls, much less offers. I wasn’t working on anything glamorous – file systems are
hardly a hot topic. My thinking was that at least I would have tried. I was fully prepared to go into
industry, and I think I would have enjoyed my time there as well.

I was super lucky to get nine interviews, and five offers. I was astounded! I had a hard time
deciding among my excellent offers, and finally chose to go to UT Austin (after a one-year
postdoc at VMware Research).

Personally, I am quite happy to have chosen this path, and after six years, I am amazed at how
well it worked out: it is a delight to work with students, I love teaching, and I love working on
research with my group.

19
But what about the cons? Yes, there is stress and responsibility, but you learn how to handle it
over time. Yes, I would have made more money if I had gone into industry. But our current
income (from both my wife3 and me) works for our family. Overall, we have an excellent quality
of life in Austin!

1.6 Choosing between an asst prof position and other jobs

This is one of the questions I get asked on a regular basis, so I figured I would explain what I
think the major differences between these jobs are. I am most familiar with (and get asked
about) researcher and developer roles in industry, so that is what I have covered. But there are
obviously other jobs such as going into government for policy work.

Choosing between two jobs always has a lot to do with the exact terms of the two jobs, location,
your goals, your financial situation, etc. So it is impossible to give generic advice. But hopefully
the following will help guide you in making your decision.

Asst prof vs researcher in an industrial lab. Let us assume you have two offers: one from an
industrial lab such as Microsoft Research, VMware Research, or IBM Research, and an offer for
an assistant professor position. I would say the big high-level difference comes down to
teaching and students. Though you will have interns and junior folks to mentor in a researcher
position, it won’t quite be the same as having your own students. In some cases, you might be
able to teach as an adjunct while being a researcher, but it won’t be part of your job per se. So if
you love teaching and mentoring students and seeing them grow, I would recommend an
academic position over one in industry.

Another thing that is meaningfully different as an industrial researcher is that for certain kinds
of research (for example, datacenters), it is hard to do meaningful research in academia. On the
other side though, there are also some kinds of research that is fundamentally impossible to do
in industry – for example, see AI ethics research that is often critical of industry.

Funding also works differently – it is easier in some cases because your manager can authorize
small-to-medium purchases easily, and harder because it might be tougher to convince the
management to give you half a million dollars for a project that doesn’t contribute directly to the
company’s bottom line.

Note that the research you can do in academia is also limited by whether you can get funding,
so it is not as “free” as it seems. However, if you want to work on a project by yourself in
academia (in other words, you don’t need to secure funding for your students) that doesn’t
require any infrastructure investment (think theoretical research), there really isn’t anything to
stop you.

3
Everyone in academia needs strong support from their partners in various forms: sometimes financial, sometimes
emotional. My academic journey would not have been this successful or this happy without strong support from my
wife and my family. We should acknowledge this and provide support to our partners and family in turn!
20
One other related point I want to make is that my friends in industrial research labs are way less
stressed and way better compensated for their time. I don’t think this would be the determining
factor, but it is something to also consider.

Asst prof vs developer in industry. Working in industry brings you a chance to make a direct
impact on the lives of your customers, be better compensated for your time (like, wayyyy better),
and have a better work-life balance. The downside is that your work is more closely tied to what
customers need, and thus can become repetitive or monotonous. Another thing to consider is
that most of these positions do not involve publishing papers, and you more or less leave the
academic world.

I’ve known excellent CS PhDs get both academic offers and a job at
Google/Amazon/Facebook/Netflix and take the latter. Several excellent professors have made
the jump from academia to industry in the last decade. So this is definitely worth considering.

I would say it boils down to the kind of work you want to do. If you want to write code, build
products, and directly affect the lives of users, the developer job is what you want. If you want to
advance the state-of-the-art and influence how the field thinks about problems in your area,
perhaps a researcher position is a better fit for you. If you want to mentor students and teach,
consider an academic position.

Additional Resources

Check out this excellent guide to post-grad jobs from Wes Weimer, Claire Le Goues, and others.
They cover finding jobs, preparing your application, interviewing, and making a decision.

1.7 A day in the life of an assistant professor

I wanted to give an idea of what daily life looks like for an assistant professor. Note that this
varies widely from person to person, so this is just one data point.

I’ll talk about my day pre-pandemic, since that is more representative than our current lifestyle. I
would typically get up around 7 AM, get my 3-year old kid ready for daycare, and drop him off at
9 AM. I would get to the dept at around 10 AM. I would typically meet with my graduate
students, and have one-on-one meetings. Meeting and brainstorming with students is generally
the best part of my work day.

Twice a week, I would teach classes at 2 PM, and I would spend about an hour prior to that
preparing for the classes. The classes would end at 3:30 PM. I would set off for home at around
4:30 PM, and pick up my son at daycare at 5 PM.

I would return to work usually after my son sleeps at around 9 PM. I would work until 11 PM and
then go to bed.

21
So on a typical day, I would work 10-5 PM + a couple of hours after dinner. If the workload is
lighter, I would skip the post-dinner work. If there is a deadline, I would work longer after dinner.

I would typically not work on the weekends. One of the nice things about family (especially
young kids) is that they force you to stop and enjoy life. It is like they know what is important :)

So this would be my life during the semesters, and we would generally travel during the summer
and winter breaks. Summer breaks we would try to visit our family in India, and winter breaks for
more touristy things.

How much time do I spend on various parts of the job?

Given that a professor's job has many different aspects, one interesting question is how much
time is spent on each aspect. Some weeks are devoted entirely to one aspect, such as research
near a deadline, so it is more useful to look at it from the viewpoint of a single year.

Research. Most of my time is spent on research. This is hardly surprising given I work at an R1
research university. This will likely not be true for folks at a teaching university. Working on
research requires both a steady time commitment (meeting with students, brainstorming,
helping them if they get stuck), and occasional sprints surrounding deadlines (writing the paper,
running last-minute experiments). About 50-60% of my time each year goes on research. Along
with teaching, this is the part of the job I enjoy the best.

Teaching. At UT Austin, I teach one course on campus every semester, alternating between
undergraduate and graduate courses. Since 2018, I have also been teaching one course every
semester as part of the online masters program. So overall, I teach two courses each semester.
I spend about 30-40% of my time each year teaching. The time spent on teaching can be
classified into class creation time (for creating a new course), class preparation time (preparing
for lectures), and grading and admin tasks. If you teach the same course more than once, you
can amortize the creation time over several semesters.

Funding. I’m a little atypical in the sense that my funding comes mainly from industry rather
than NSF or other government agencies. Thanks to the generous UTCS startup + industry
funding, I could afford to write fewer proposals (and get fewer grants). When I started as an
assistant professor, I spent a lot of time and effort on the NSF CAREER. Once I got that though, I
didn’t spend a lot of time writing grants since I was getting funding through other means. This
might change post tenure though. I spend about 10% of my time each year (about a month) on
funding-related activities: either writing grant proposals, or visiting companies and giving talks.

Service. When I started out as an assistant professor, I signed up for too much service,
sometimes serving on 5+ program committees in a single year. I’ve since learned to cut back on
service and only take on a few program committees and other organizational tasks a year.
Service should take up 5-10% of your overall time, but it took up 10-20% of my time when I
started.

Mentoring. Mentoring PhD students is an important part of being a professor. It is hard to think
about time taken especially for mentoring, since it is usually woven together with research

22
meetings. Usually I meet with each of my students one-on-one at least once a week, and that
serves as time for both research and mentoring. So I would subsume this into the time taken for
research (which is the majority of where my time goes).

If you notice, the percentages sometimes add up to more than 100% when I’m having a year
with a lot of service or teaching a course for the first time. At times, the job is definitely more
than 40 hours a week (closer to 50-60). It is part of my conviction that professors are grossly
underpaid, and academia (and society in general) pays for this mistake when excellent
professors leave academia to go to industry.

Additional Resources

Prof. Casey Fiesler at the University of Colorado Boulder has a great set of videos about the
academic job, applying for such jobs, etc. Check out this particular video about the prof job at
an R1 university.

Summary
This chapter should give you a good grasp on what is involved in the professor job at an R1
university. We talked about the financial aspects of the job, how consulting works, and pros and
cons of the job. I described why I chose this job, and what a typical day looks like for me.

Now that we have covered the overall picture, it is time to dive into each of the responsibilities of
a professor. In the next chapter, I will talk about finding and recruiting good students, and
mentoring them to be successful.

23
Chapter 2: Students
Now that you know what the assistant professor job is like, let us talk about PhD students. Your
career will depend on your PhD students: their success is your success, and a large part of the
job is about training PhD students to be successful, independent researchers. So how do you
find, recruit, and mentor PhD students?

Let us first talk about what the admission process looks like.

2.1 Recruiting students: the admission process

If you are reading this book, it is likely you have gone through this process as a graduate
student. But let me supply some more details as to what recruiting students is like from the
perspective of a professor. Note that I’m talking about PhD admissions here: masters
admissions are a different beast.

The application process. The process begins by the department putting out a webpage where
students can apply. This will collect details such as name, educational qualifications, statement
of research/purpose, transcripts, and scores from GRE and TOEFL. Students have to provide
contacts for three people who will write letters of recommendation for the students. At some
schools, the student might have to apply twice: once to the dept, and once to the graduate
school/university.

At UT Austin, the student can specify up-to three professors you want to work with; this is useful
since it allows professors to focus on the students who are interested in working on them. At
some places, you just mention the professors you want to work with in your research statement
(making it much harder for the professors to find students who want to work with them).

There might be a fee for applying. At some universities, this revenue is used to pay the support
staff who handle applications. This fee might be waived in certain situations; you should always
email and ask.

The application process generally runs from Sept to Dec each year, for admission for Fall
(Aug/Sep) the following year. At some schools, the application process runs longer (until Mar).
Some schools also admit students in Spring (Jan/Feb) with an earlier application process.

For professors who are looking to admit students, it helps students a lot if you are explicit about
1) whether you are recruiting students, 2) what you are looking for in prospective students, and
3) what you offer. Inspired by Philip Guo (UC San Diego), I’ve written something like this here.

Other than this, professors don’t have to do much for the application process. Most of the work
is handled by the awesome administrative staff in the department.

24
The evaluation process. How students are evaluated and admitted varies widely from
department to department. Generally, an admissions committee will evaluate students and
decide who to admit. Most of the committee members will be professors, though some
students might be included as well.

The committee is generally looking to ascertain two things: 1) does this student have the
intellectual chops and the persistence to do a PhD, and 2) is there a fit between the student and
one or more faculty at the dept. A PhD is an apprenticeship – you will be working with a specific
professor, so it is not enough to be academically good; if no professor wants to work on the
topic you are pursuing, admitting you is not a good idea. So finding a good match is important.

The committee determines intellectual chops and persistence mainly by looking at grades and
prior research experience. If the student has done research with a professor and the professor
says in their letter “I’ve trained a lot of PhDs and I think this student has what it takes”, that
counts for a lot. In the absence of research experience, doing well in courses is a good indicator
(though not the best indicator, the committee doesn’t have much else to depend on in this
situation).

How does the committee determine student-prof fit? The committee leaves it to the profs
themselves. At most departments, as a professor, you will be able to examine all PhD
applications. You can contact the students you are interested in, and interview them. You can
then tell the admissions committee “I am interested in students X, Y, and Z”.

At some universities, a professor has to commit funding to indicate interest in a student. So if a

professor is interested in a student, they will have to commit to paying them for one year (see
Chapter 3 for how much this can cost). This prevents professors from expressing interest in
students without having sufficient funding. At some schools, professors combat this by pooling
their resources and saying “Professors X, Y, and Z are interested in this student”.

UT Austin does not require committing funding to express interest in students. This is a more
flexible scheme, though it does lead to professors taking on students without first having the
funding to pay them. UT Austin guarantees funding for all PhD students for five years via
teaching assistantships; otherwise this could be bad for students. Many schools (though not
all) will guarantee funding via TAs for some number of years.

How does the committee determine how many students to admit? The committee will ask
profs in the dept, and they will each say something like “I am recruiting 1 student this year”. The
committee then has a sense of “demand” in each area: for example, it knows we are looking to
hire four students in systems this year. If the demand is N students, the committee can’t admit
exactly N students, since some students will choose to go to other schools for their PhD.

25
The committee will look at historical yield, the percentage of students offered admissions who
accepted the offers, to decide how many offers to make. For example, at many universities, the
yield is around 30%; the committee will then make around 3*N number of offers. Sometimes
this backfires with a lot of students accepting the offers, what is called a “success disaster”; it
just means the dept won’t be admitting as many students in the next few years.

When you join as a professor, ask and figure out all these details – it will determine how
aggressive or conservative you need to be when indicating interest in students. For example, if
you can only commit funding for one student, it doesn’t make sense to interview more than
three or four students for that role.

2.2 Attracting students to your group

Alright, we have covered the process of how students apply and get admitted. But how do you
get students to apply in the first place?

The more famous and highly-ranked your university is, the less of a problem this is in general.
But even so, if you care about diversity and inclusion, you might need to make an effort to get
students from those groups to apply.

Let’s look at this from the viewpoint of students. What makes an attractive destination for their
PhD? Students want to go to a university where 1) they can work on their topic of interest, 2)
with a stable funding situation, 3) with good job prospects, 4) that seems fun!

For international students, rankings matter a great deal as they are not able to visit the
universities in person before making a decision. So students typically end up going to the most
highly-ranked university among their options.

This can be both good and bad: higher ranked universities typically have more money, attract
stronger students and faculty, and have a stronger brand that can help with job hunts and more.
However, international applications tend to place too much emphasis on ranking, thinking that
there is a lot of difference between ranks 14 and 17 (there is not). But in any case, professors
should be aware that this is happening.

The most concrete thing you can do to attract students is do good research and build a good
brand. The good research part is obvious: if you publish at good conferences and have
real-world impact, that is going to be attractive to students who wish to do the same. The brand
part is especially important if your university as such doesn’t have a great brand. You need to
get the word out about your research group, especially if it is a new group at your
department/university. For example, if your department has not had any faculty working in HCI
for decades, HCI applicants may skip over it; you need to make sure potential applicants are
aware of your new group.

26
By “build a good brand”, I am not talking about anything complicated. You should spread the
word about your research, both your papers and your broader vision. Talk about what kind of
group culture you want, what you expect from students, and what you look for. Talk about the
story behind the paper: the struggles, the different directions you explored before coming up
with this, etc. Basically, talk about what life would be like in your research group.

Doing outreach. I recommend doing outreach to encourage students to apply. I did outreach at a
number of Indian universities (including my alma mater). Doing outreach in post-COVID times is
easier than ever: simply email someone you know at the university about giving a Zoom talk. I’ve
never heard of anyone declining a talk, especially a Zoom talk. The talk should present your
vision for your research area, the research you have already done, and some of the research you
are working on/planning to do. Don’t make the talk too technical: it should be similar to a job
talk – the overall picture is what matters. You should talk about what you offer to your students
(they may not know that PhD students get a stipend!) and what you expect from them. If the talk
is good, students will remember your group and apply, and faculty at the institution will
recommend their students to apply in later years.

Developing pipelines. Some students will apply to your department just because of the
university/department name. But you also want to develop “pipelines” of students from places
you trust, where they apply specifically to work with you. Your undergrad alma mater is a good
candidate for this. But you can also find other places. For example, I like to recruit from
Microsoft Research India: they have a Research Fellow program where students can do
research alongside experienced researchers for two years, and then apply for grad school. It is
worth building a relationship with such places so that students know you and apply to work with
you.

Looking out for diverse talent. The location of your university influences the demographics of
the applicant pool (except international students). For example, some universities might get a
lot of Hispanic applicants due to their location. If you want more diversity in applicants, you
have to reach out to them. For example, giving talks at Historically Black Colleges and
Universities (HCBUs) is a good way to encourage African Americans to apply to your research
group. Diversity leads to different perspectives, which then leads to more creativity and
innovation.

2.3 Identifying good students

Once students are applying for PhD admissions to join your research group, how do you identify
students who are a good match for you?

It is crucial to pick strong students who work well with you, especially at the start. Your first
students form the core of your group: they will mentor other students, your pre-tenure papers
will likely be their work, and they will strongly influence the culture of the group. So pick your first
one or two students carefully.

27
There are three situations where you have to evaluate students: when they have already been
admitted to your university (new PhD students), when they are PhD applicants, and when they
are potential PhD applicants (who are emailing you about wanting to work with you). We will
discuss each of these situations.

Evaluating PhD students who are already admitted. Let me first talk about a scenario where you
have joined the department in the fall, so you don’t have any students, but there are a number of
newly admitted PhD students who also started in the fall. Perhaps some of them don’t have
advisors yet, and they are interested in working with you.

There is a temptation to quickly hire anyone who will work with you: you have ideas, you have
funding (thanks to your start-up package), what you lack is time to execute those ideas. But
resist this temptation: students who aren’t a good match will suck up a lot of your time, and will
be a net negative. If your first student is someone who doesn’t work well with others, other
students will not want to work with you.

In my own case, I didn’t know this at the outset, and basically got super lucky. This is why I say
my success has depended a lot on luck at multiple points – of course I worked hard, but it is not
hard to see that I was also lucky to get good students initially.

The best way to identify if there is a match between you and a student is to simply work with
them for a while. Typically, faculty work with incoming PhD students for a semester or a quarter;
if there is a match, they formally take them on as students from the next semester, funding
them as a research assistant.

I recommend this “evaluation” period be shorter than a semester. In case there isn’t a match, it
is a waste of time for both you and the student. It is particularly expensive for the student, who
can’t afford to spend several semesters rotating with different professors before finding their
advisor.

I suggest making a decision in about one month. You might think, isn’t one month too short?
Remember that you are not looking for the “maybe” cases here: you are looking for someone
who is a clear and obvious match, who fits well with you and your style of research from the
get-go. You can tell if you are working with such a person within a month.

During this month, work closely with the student: meet with them twice a week (or whatever
cadence works for you). Start by giving them concrete tasks: don’t give them a task like “here is
this open problem, think about it and come back”. Give them tasks where you know the solution,
so you know how long it should take to do something. Ideally, the task should take about a week
to complete; that way, you know if they are making progress at every meeting. At the end of one
month, if you don’t feel strongly that this will work out, tell the student politely, let them work
with some other professor.

28
One failure mode here is to feel bad for the student, and delay the separation point. Remember,
if it is not ultimately going to work out, it is better for both you and the student to end the
relationship as soon as possible. Don’t feel bad here: it is better if the student works with
someone who is a good match for them.

Another failure mode is being unable to make up your mind about the student. Perhaps after a
month, you feel like you are not sure it will work out. My advice in this situation is to find a
different student. The advantage of having a high bar initially is that you can tell pretty quickly –
if you find yourself not feeling sure, then this is not the student you are looking for.

There is a chance that you don’t find a student for one or two years – I know successful
professors who went through this. While looking for your first PhD student, make sure to work
with undergraduate students and masters students who are doing research. If there are other
faculty in your area, co-advise their students with them if they are working on something
relevant to your interests. But continue to have a high bar for your first few students on whom
you will spend your time and money.

I want to note that this is not the only way to find good students. This is highly subjective, and
varies from professor to professor. There is merely one data point, based on what worked for
me.

Evaluating PhD applicants. Another situation is that you are evaluating PhD applicants who will
be admitted for the next year. This is a much harder problem, since you can’t really work with
students to figure out if they are a match. However, remember that you are only looking for
students who could potentially work with you: it may not work out after a student is admitted
and starts working with you. So you are looking for the students who, if they are admitted, you
would want to work with and evaluate.

Examine the applications of students who say they want to work with you. You might have
subjective criteria you use to determine a match: for example, I look for experience building
software (such as Google Summer of Code), since this is a necessary skill in systems research.
If you have done outreach before and you know and trust some institutions (such as your alma
mater), keep an eye out for that too. Remember that not all talented students will have had the
opportunity to do research, and that grades are unreliable for international students.

Make a short list of the students who seem promising. Email the students and schedule a Zoom
call with them. In these calls, I try to get a feel for the students and whether they would be a
good match for my group. I find that talking to the students for a short time (~30 min) is
extremely useful in figuring out if they would be a match. The calls are enough for me to find the
few students I’m interested in working with each year.

Evaluating potential PhD applicants. Once you start as a professor, you will regularly receive
email from students who want to work with you. You can ignore these emails, or just ask them

29
to apply via the regular application process. But sometimes these emails allow you to find
diamonds in the rough.

The essential difference between PhD applicants and potential PhD applicants is time. If the
students email early enough, there is still time to work with them remotely to figure out if there
is a match. Here is what I do. This doesn’t work for non-systems PhDs, so it is not a general
technique, but you might be able to adapt it.

Most of my projects are open source on Github. For each project, I create small issues on
Github for small, simple additions/updates that I want to do over time. These are simple enough
for an undergraduate student to tackle. So when a student emails me saying they are interested
in working with me, I simply point them to Github and ask them to take a shot at any of the
issues.

This step is amazingly effective: more than 95% of students who email vanish at this point. The
remaining 5% tackle the Github issues, and about 1% actually ask meaningful questions about
what they have done so far and what to do next. Those 1% are worth mentoring, and I have
found a couple of strong PhD students in this manner. I always write letters of recommendation
for all the students who work with me in this manner remotely, regardless of whether they are a
good match for my group.

The drawback of this approach is that you are asking students to do a significant amount of
work on top of their regular undergrad workload. Not all talented students would be able to take
the time to do this. Unfortunately, I know of no other way to find the “diamonds in the rough”.

So now we have talked about the admission process, how to attract students to your group, and
how to evaluate students who apply. Let us next discuss how to manage students.

2.4 Managing students

Most professors don’t have any formal training in managing people when they start as a
professor. I didn’t when I started. So if you are reading this as a graduate student, know that
there exist workshops that you can attend to get some of the training you need!

A big part of being a professor is managing students. So what does it actually mean to
“manage” a student? It is a vague term, but roughly it means that you are responsible for:
1) Your student’s overall research career until they graduate. You train them on how to do
research and how to communicate it. You help them pick suitable problems to solve. You
help publicize their work. You nominate them for the appropriate awards, and help them
get a job by introducing them to your network.
2) Your student’s research progress. This is more like “management in the small” where you
help students who are struggling find a way to move forward, or when appropriate, figure
out when to drop a project.

30
 3) Your student’s personal well-being (to an extent). An enormous part of your student’s life
is going to center around their PhD and their research. Thus, managing a student
includes things like motivating them, cheering them up when they face rejection, and
encouraging them to take care of themselves. This does not mean you are their therapist,
but it does mean you might have to encourage them to see one at times.

An advisor bears significant responsibility towards their students: in a sense, they are entrusting
their academic career to you. You should make sure not to abuse that trust.

A lot of what I’ll say in this subsection will seem obvious to experienced managers. But it wasn’t
obvious to me when I started, so I think it is worth laying it out.

An important thing to remember when you hire your first PhD student: don’t expect them to be
as productive as you are! It is unfair to compare a first-year PhD student (with perhaps no prior
research experience) to your trained, experienced self. It is sometimes hard to imagine us back
as we were in our first year when we didn’t know anything. So be patient with your students!

Work hours and being professional. This is pretty basic: your relationship with your student is a
professional one. You are their employer and their manager. You have a lot of power over their
career: don’t abuse it by asking them to do non-professional things. You can be friends with your
students, but you still need to maintain that boundary. Barring deadlines, you should not be
expecting them to be available for meetings in non-work hours and the weekends. Don’t
intertwine your personal life with their personal lives.

The basic setup. Many professors meet with their students one-on-one for about 30 minutes to
an hour each week. In addition to this, some professors also have a group meeting that is used
to communicate what everyone is working on to the rest of the group. This is especially useful if
your students are working from different places. Use a spreadsheet, or whatever organizational
software you fancy, to keep track of what each student is working on: the papers that are
submitted, where they are in their PhD, and the next milestone/deadline they are working
towards. This is easy to remember when you have one student; I have six students currently, and
using tools to keep track definitely helps!

Onboarding new students. When you accept a new student into your group, there are a number
of things you should do:
- Add them to the group slack/mailing list
- Inform your admin that you will be funding them for the semester (and hopefully beyond)
- Sign any paperwork required that denotes you as the student’s advisor
- Introduce the new student in person to the rest of the group. Go out for a group dinner
and help the new student make some connections in your group
- Set up a weekly meeting time for them
- Talk to them about your expectations for the first semester – this could just be reading
papers as they take required classes, but whatever it is, make sure you are on the same
page.
31
 - If they are weekly or other semi-regular events that you would like them to attend, talk to
them about it
- Come up with a first task for them to work on. You could ask them to read a paper, or set
up some software, anything that will aid their research. But give them a concrete task.

The first semester, and the first year, can be particularly stressful for students. You can help
them navigate it by introducing them to folks and events, and being clear and upfront about your
expectations.

Though this typically doesn’t fall under “onboarding”, when a student joins your group, you are
responsible for giving the student a problem to work on. Don’t expect a first-year PhD student to
figure out what to work on: they lack the required context and knowledge base to do so. One
way to look at a PhD (which I’ve heard from many sources) is as a progression: first project, you
give the student the problem, you tell how to solve it; second project, you give the student the
problem, they figure out how to solve it; third project, the student comes up with the problem
and the solution. So the first project of a PhD student, you need a problem that they could work
on. Typically, a newly-hired assistant professor is bursting with ideas, so this shouldn’t be a
problem. But remember to give them enough guidance to solve the problem.

You should tailor the project to the unique skills of the student. This is especially important for
the first project where the student doesn’t have confidence yet. If a student is strong in building
systems, don’t give them a problem involving theory (or vice versa). Play to their strengths, and
make up the gaps with your own skills or other students.

Management style. Management styles vary widely among professors. Some professors are
hands-off, meeting with their students once a month. I have even heard of professors who only
meet with their students once a semester, typically to sign forms and get a high-level update.
Other professors are super hands-on, basically working with their students on a daily basis.
Depending on the working style and personality of the student and the advisor, a lot of different
management styles can be a good fit.

Generally though, I find that with junior students, being hands-on and communicating frequently
helps a lot. Junior students need a lot of hand-holding, and especially for the first one or two
students, they are getting 100% of their mentoring from you. Once students are more senior,
they can help mentor new junior students. But it is important to invest the time in training the
first few students.

Two key things helped me in managing my students: setting expectations, and communicating
frequently. Students can feel unhappy during grad school due to two issues: the lack of
structure and clear goals, and lack of communication and feedback. A student can feel adrift at
sea: no clear goal to work towards, or trying to tackle a problem that seems too hard, unsure of
how to do the next step. This problem is further compounded if the student infrequently meets
with their advisor, leading to a feeling of being alone and being lost.

32
In my group, we combat this through “over-communication”. I basically tell my students to
contact me whenever they have something to discuss, they don’t need to wait for our weekly
meeting. I chat with my students several times a week when I see them in the dept, and if my
door is open, they are welcome to come in and chat.

Back in the pre-pandemic days, I would usually start my day by going around their cubicles and
chatting with the students. They would tell me if they are blocked on anything, and if it is
something small, I unblock them right away. Otherwise, we plan to meet later in the day. In this
way, I’m always aware if a student is struggling.

When a student is having trouble figuring out the next step, we meet and brainstorm. I help the
student with decomposing the problem into smaller parts, and figuring out what the next step to
take is. Oftentimes, it is the act of meeting itself that is important: in this environment, the
student figures out what to do by themselves. Often, my role is to provide the “scaffolding” for
the student to get to the answer themselves; sometimes I provide technical input (typically
when the student is more junior) but the non-technical help is more significant.

This style of working doesn’t fit all students though: some students hate the constant
interruption and meeting, so these students are not a good fit for my group. This is the kind of
stuff that is hard to figure out by reading an application.

Setting expectations. One of the most important things when a student joins (or even before
they join) is to set expectations. Tell them how long a typical PhD takes, and tell them about the
milestones they are expected to hit (and when they are supposed to hit them). Ask them about
their post-PhD goals, and tell them how you will help them attain those goals. At an individual
project level, set the rules of how long you would tackle the problem together before declaring
“too hard right now”, shelving it, and working on something else. If you are evaluating a student,
tell them clearly what you expect at the end of the evaluation period. If you are initiating a big
joint project, talk to the student about their role in the project, and what they can expect to get
out of it. For example, for any project with multiple students, I make it clear who the first author
is before we start the project. I talk about whose dissertation the work is going into. Setting
expectations upfront helps side-step a lot of problems down the line.

Handling different kinds of students. Your students are going to have a variety of working styles
and communication styles. If you are advising a student, the two of you have decided that you
have compatible working styles. Even then, there is a lot of room of variety. Some students
might work 9-5 pm, others might be night owls. Some students may prefer to be in the lab all the
time, others will only come in for meetings. Students might also have different goals after their
PhDs.

For each student, you have to tailor how you mentor them based on how they work and
communicate. Mandating one style for the lab will not work. You will have to figure out what you
should ask of them so that they make progress without impeding their natural style. For

33
example, asking a person who mainly works at night to come to the lab early in the morning isn’t
going to work.

At the end of the day, what matters is whether the student is productive. If the student is
working in a team, that introduces additional constraints, but you should generally let the
students have as much freedom as possible within those constraints.

Motivating students. A crucial aspect of being a professor is motivating your students,

especially through inevitable rejection. This also applies to yourself – if you are becoming a
professor, you have been pretty successful in academia; perhaps you didn’t get a lot of
rejections coming your way. But you will in the future, irrespective of how good you are; I’ve seen
successful, award-winning professors have their papers and grants rejected (a lot).

I recommend a “rejection routine”. When the decision comes in, if it is a reject, don’t read the
reviews: there is nothing urgent you can do anyway. Go out with your students and get an ice
cream, let the students rant, and get it out of their system. You can return to the reviews when
you are ready, and see if you can use the reviews to strengthen the work.

When you meet with your students, appreciate them when they have done good work.
Appreciate progress in their work. Be generous with compliments and praise. Sometimes as
professors, we don’t really realize the enormous value a kind word has to our students. When a
student is working through a long project that takes months or years, it is important that they
receive regular positive feedback.

What to do when expectations aren’t met. Sometimes one of your students may be struggling
with their project. It might have been a while since they made any significant progress in their
project. The best thing to do here is to involve yourself as much as you can in their project, and
help them past the hurdle. One of my mentors always ensures a student is un-blocked before
they exit their office; they work on the problem together with the student until then.

Sometimes, though, despite your help, the student struggles. This happens for various reasons
– working style mismatch, personality mismatch, something happens in their personal life, etc.
When this happens, it is important to be upfront about it, and see if you can fix it. Always set a
time limit for the fix (“let us see if things improve in a semester”). Sometimes the cause is
fundamental and not fixable: the student’s working style just doesn’t work with you and your
group. While the hope is to figure these things out early on, sometimes they surface later. When
this happens, you should sit down with the student and explain that you are ending the
relationship; you should continue to pay the student for the remainder of the semester, with the
understanding they find different funding from the next semester. Don’t get into “why”, it is rarely
productive; politely but firmly explain that you are not a good match, and that you wish them
well. If they are dropping out of the PhD and going into industry, help them with the job hunt.

34
2.5 Setting group culture
Group culture can be an important part of whether your lab is a thriving, collaborative place. I’ve
heard, first-hand, of highly successful groups in top universities where students don’t voice their
ideas because other students will steal them.

An atmosphere like this can be avoided by setting expectations upfront (“who is first author on
this project?”) and acting quickly when you see a violation. For example, let us say a student
makes a misogynistic comment. A quick, sharp “We don’t make remarks like that here” will be
sufficient. If this keeps happening, you might need to have a talk with the student. But most
students are sensitive to stuff like this, and will react quickly. Over time, the culture is embodied
in your students themselves, and they will pass it to new students joining the group.

Students will feel safe in collaborations if they know their own ideas are safe and that their
end-goals will be met. Have the difficult conversations about authorship and whose dissertation
work goes into, up-front. Don’t wait until the work is done: a student thinking they will be first
author will be understandably upset if they learn they will not be.

Lay the platform for students in your group getting to know each other by taking them out to
shared events such as dinner. Make sure new additions to the group are always included in
team events.

Overall, if you take an active role in shaping the group culture in the initial years, it will self
propagate as you add new students.

2.6 Frequently Asked Questions

Q: How can I fairly and equitably evaluate candidates?
A: This is a hard question. I like the saying, “Talent is uniformly distributed, but opportunity is
not”. There are students out there who have the raw talent, but are unfortunately never exposed
to research, and therefore never even consider a research career. These students will not even
apply to grad school, so the solution here is outreach. Give talks in places that typically do not
engage in research so that students know about this career option.

Once they do apply though, how do you find talented students who have simply not had the
opportunity to do research? In systems, I tend to look for adjacent skills that would help in doing
systems research, such as being a good systems-builder, or being a good communicator. I’m
sure other areas can come up with similar adjacent skills to look for. As I described in Section
2.3 (evaluating potential phd applicants), asking applicants to do a small but simple
research-related task is effective in finding candidates who have the initiative and the
persistence to succeed.

Note that when you are dealing with international students, grades don’t count for a lot.
Depending on the university and the department, a converted 3.5 GPA might be terrible or
35
amazing. It is hard to tell without context. Similarly, GRE scores tend to measure “being able to
prepare for the GRE” and not much else (I didn’t have a great GRE score myself).

Q: What can I ask students during the chat to figure out if there is a match?
A: Remember that you are trying to figure out if there is a match between you and the student,
among multiple axes: research interests, personality, and work style. I usually find the answers
to all these questions just by asking the student to talk about some research or project they did,
and follow up with questions about the work. There is no checklist of questions per se that I
use. Usually by the end of the interview, I can tell with some certainty whether there is a match.

Q: What kind of info should I make available to PhD applicants?

A: I talk about this in Section 2.1: Apart from specifying whether you are looking for students,
please have a public web page talking about what you are looking for in prospective students
(what kind of skills are you looking for?), and what you offer (for example, do you encourage
internships?). Philip Guo (UC San Diego) originally came up with this idea – I’ve written
something like this here.

Q: What channels should I use to recruit?

A: I recommend using resources such as the CS Slack, Twitter (if it still exists when you read
this) and other social media. Outreach at different undergraduate universities also pays off
nicely.

Acknowledgements
Thanks to Ryan Marcus for contributing some of the questions this chapter focuses on. Some
of this chapter was inspired by the Mochary Method Curriculum for founders.

Summary
This chapter covered one of the core responsibilities of a professor: attracting, mentoring, and
managing students. We went over some of the tricky details in how you evaluate students at
different stages of the application process.

Next, we will talk about obtaining funding for your research group. This is another role that you
have to become competent in as a professor.

36
Chapter 3: Funding
Let’s talk about what, for many professors, is the least favorite part of the job: making sure you
have enough money to keep the lights on in the lab. Essentially, a professor applies for grants
(money from the government, industry, or other charitable foundations) by writing a proposal as
the Principal Investigator (PI). The funding agency reviews the proposal. If they decide to fund it,
they send the money to your university. The university then takes its cut of the funding
(essentially to pay for common services like buildings, admins, etc) and gives you the rest. You
then spend it on your summer salary, the salaries of your students, equipment, travel,
registration costs and other expenses associated with your research.

There are some nuances about the different types of funding, and how you can spend different
kinds of funding. So let us get into the details. Unfortunately, these details tend to vary a lot by
country, so everything I’m talking about in this chapter applies only to the US.

3.1 An example budget

Let me walk you through an example budget. Let us say you are writing a proposal for $100,000
USD (100K USD). The budget will talk about how you plan to spend this money. The following
table (Table 1) shows the example budget.

Budget
Graduate Research Assistant - 1
Total Salary 40,000
Fringe Benefits 12,000
Total Salary and Fringe Benefits 52,000
Total Materials & Supplies 0
International travel to a conference 2500
Domestic travel to a conference 1481
Total Travel Costs 3,981
Tuition for Fall or Spring 4,746
Tuition for Summer 1,778
GRA Tuition for 1 year 11,270
Total Direct Costs 67,251
Modified Total Direct Cost (MTDC) 55,981
Indirect Costs 58.5% of MTDC 32,749
Total Cost (Direct + indirect costs) 100,000

37
I will explain the various parts of this example budget. I will be talking about terms such as
direct costs and indirect costs as defined by the University of Texas – I believe it is similar at
most other US universities, but there might be differences. Please check with your university to
make sure.

Fringe. Let us say you want to support a student on this grant. In addition to paying for the
student’s salary, you also have to pay for non-wage expenses: things like health insurance, social
security, and medicare. This is usually expressed as a percentage of the student’s salary. At the
University of Texas at Austin, this was 30% in 2022 (For a comparison point, it was 32% at
Princeton). The stipend of a graduate research assistant (GRA) at UT is about $30K USD for the
fall and spring semesters ($40K USD including summer). So salary and fringe for a GRA for a
year is $52K.

You also have to pay the tuition of the GRA, and at UT, this is about $11.2K USD for a year. To
keep things simple, I am using a single number for the tuition, but this depends on factors such
as whether you are in-state or out-of-state, your progress in the PhD program (if you are
all-but-dissertation, your tuition is lower), and whether you are a US citizen or not. It also
depends on the fraction of time spent as GRA/TA: if this drops below 20 hours a week
(considered full-time for students), tuition might have to be charged at the out-of-state rate. If
the student is only doing 5 hours a week as GRA, it might also impact how much of their tuition
can be covered by the grant; at 5 hours a week, the student might not be eligible for insurance. I
highly recommend talking to your admin if you are accounting for less than 20 hours a week for
your student on the project. Some schools also calculate tuition as a % of the stipend.

So the total cost for supporting a GRA for one year is $63.2K USD. While you are paying $63.2K
for each student for a year, the student's take-home pay is only the stipend part (minus taxes
that are withheld): $35.2K assuming a 12% tax rate.

Note that fringe also applies to you (the PI) when you pay yourself out of your grants for your
summer months.

Direct costs. The direct cost part of your budget includes salaries, fringe benefits, travel, and
equipment and supplies that contribute directly towards performing the proposed research.
Direct costs are exclusive to each proposal: a piece of equipment can only be a direct cost on
one proposal, and a GRA can be assigned only to one proposal at a given time. In our example,
let us assume the only direct costs are supporting the student and paying for the student’s
travel (including registration, transportation, lodging, and food) for one international and one
domestic conference.

Modified Total Direct Costs (MTDC). MTDC is a portion of the total direct costs used for
calculating the indirect costs. It is the direct costs minus tuition, fellowships, equipment, and
capital expenditures. Travel also counts as MTDC. In this budget, we are not spending anything
38
on materials and supplies in this budget, but if we did, it would be part of MTDC. Check with
your university to determine exactly what is counted as MTDC – this can be surprisingly tricky.
For example, you might expect computers to not count towards MTDC as it is “equipment”, but it
may be counted as a “supply” if the university is unable to amortize its cost. In the example
above, total direct costs is $67.2K USD (63.2K for one GRA plus 4K for travel), but MTDC is only
$56K (leaving out the $11.2K for tuition).

Indirect costs. The university takes a portion of all the money you receive, to pay for centralized
functions such as building costs, utilities, admin, and libraries. This is termed as the “indirect”
cost, usually expressed as a percentage of the MTDC. At most universities in the US, indirect
costs are more than 50% of the MTDC. At UT, it is 58.5% (Mike Freedman reports it is 63% at
Princeton). Thus, in our budget, we would have .585 * 56K = $32.7K added as indirect costs. As
a result, it would cost 67.2 + 32.7 = $100K USD for supporting a single GRA and some travel on
this grant. Note that even though the 58.5% number is scary, you are only paying 33% of your
budget for indirect costs; in other words, you don’t lose 58.5% of the total budget to indirect
costs.

As you can see from the preceding table, the $100K USD budget only covers one student for one
year. At some universities, a GRA costs more, requiring more funding.

This should give you a picture of how expensive it is to maintain large groups! If you had ten
students, you would need to pull in a million dollars each year to fund your students.

It is useful to think of your budget in terms of student-years. $400K is simply two students for
two years. A dissertation, roughly five student years, is $500K. If you get the NSF CAREER grant,
you would typically be able to support one dissertation. This helps determine the scope of your
proposal: if you need five students to work on it for two years, a grant that provides $250K will
not be a good match.

Summer months in your budget. PIs at most R1 universities in the US are expected to fund
themselves via grants in the summer. When you are drawing up your budget, you are expected
to estimate how much of your summer months you will dedicate to this project. This is taken as
an indication of PI commitment at the funding agency – you don’t want to write a budget where
your estimated time involvement is zero. So if you have one NSF grant where you estimate your
time commitment to be one month, you can only put in one more month for another NSF grant
proposal (since NSF limits total PI compensation to be two months per year across all NSF
grants).

3.2 Types of grants

Broadly speaking, there are four different kinds of grants or funding, each with different
restrictions on how you can use the money. The differences come down to what you can spend
the money on, how long you can keep the money, and what the overhead is.

39
We can term the four kinds of funding as grants, contracts, university funding, and gift funding.
Let me go over each type of funding.

Grants. The National Science Foundation (NSF), the Defense Advanced Research Projects
Agency (DARPA), the Army, the Navy, and the Air Force all fund research related to computer
science. These grants are competitive – NSF has a 25% acceptance rate (though this varies by
the directorate), and I imagine the other programs are similar.

This money can typically only be used for funding students, yourself, travel, and equipment
purchases that you had already put in the budget. For example, if you suddenly want to
purchase equipment not in the original budget, it needs to be approved by the Office of
Sponsored Research at your university (which makes sure you are spending the money in the
right way). You cannot use government funding to pay for things like a meal with your students
during the semester (meals during travel is allowed).

Government funding has strict timelines by which the money has to be spent. Typically, the
money is released on a yearly basis. You have to show that you have used the previous
installment to get the next round of funding. This sometimes results in professors making last
minute purchases to exhaust their grant!

In some cases, you can ask your funding program manager for a no-cost extension to your
grant. This typically means that you don’t need extra money, but you want extra time to use the
money. This is usually considered on a case-by-case basis, and I would not recommend
counting on getting a no-cost extension.

Indirect costs are levied on government funding (more generally, on all funding that goes
through the Office of Sponsored Research). The example budget I showed in the previous
section assumes funding is via sponsored research.

Grant proposals typically have reporting requirements (for example, you have to submit a report
once a year for an NSF grant). However, grants typically do not expect any concrete deliverables
to be provided at the end of the funding period. Essentially, the grant allows you to carry out the
proposed research: the act of publishing itself is the desired outcome. As a result, grants have a
strong alignment between the PIs and the funding agency: they both just want to get research
done.

Contracts. The second kind of funding is via contracts. Contract funding comes with an explicit
contract you must fulfill to keep and obtain more funding. For example, there may be quarterly
or monthly reports, or on-site visits by the sponsor. Contract funding typically also has
deliverables that are expected at regular intervals during the funding period. One example of
contract funding is DARPA funding, which comes with a number of deliverables and reporting
requirements. Contract funding is less aligned with the PIs than grant funding: apart from doing
the core research, the PI’s students will have to spend time on creating the deliverables, and the
PI herself has to spend a significant amount of time on reports and meetings.
40
NSF vs DARPA. NSF is a good example of grant funding, while DARPA is a good example of
contract funding. Both these agencies fund a lot of computer science research. DARPA requires
deliverables and reports to be produced regularly during the grant duration. NSF only requires a
report to be submitted for each year of the grant duration. You can apply for NSF funds as
someone on a visa, while DARPA and other similar government funding programs might require
you to be a US citizen or a permanent resident.

University funding. Your university, or your college, or your department may choose to give you
funding. Typically when you start, you get a “startup” package that allows you to kickstart your
research group. University funding is “internal” funding in a way, and as such has fewer
restrictions than government funding. But this really depends on the specific university. I know
universities where startup funding cannot be used for covering summer salary of PIs. Apart
from startup funding, your university might have internal competitions to award “seed” funding
(typically under 50K) that you can then use to get preliminary results and later apply for
traditional grants.

University funding is also associated with timelines. For example, startup funding usually has to
be spent before you get tenure. However, university funding deadlines might be a little more
flexible than grant or contract funding deadlines.

University funding typically does not have associated indirect costs (as it is already university
money). You still have to pay fringe support for your students and your own summer salary.

An interesting sub-category of university funding are Research Centers. A research center is

funded by sponsors to carry out a certain mission over a span of around 5 years. The research
center has a program director who is responsible for allocating the funding. PIs at the university
apply for funding from the research center (a sort of internal grant), and are allocated funding.
Obtaining funding in this manner is easier than competing for an external grant. Particularly for
assistant professors who are just starting, research centers can provide seed funding before the
first grants are obtained. Check out the ADA center at Michigan for a good example.

Gift funding or unrestricted research funding. Companies or charitable foundations may

choose to fund your research in the form of “gifts” or unrestricted research grants. This is the
best kind of funding, for three reasons: indirect costs are typically not levied on these gifts, there
is no associated timeline by when you should spend them, and you can use them for a broader
range of things than university and grant/contract funding.

Indirect costs are not levied because the gift-giver can specify that it cannot be used for indirect
costs (the university has to honor this). Since it is an unrestricted gift, there are no limits on
what you can spend them on, or by when you should spend them. This doesn’t mean you can
spend the money on personal expenses. All expenses still have to be approved by the university,
and should be used for research-related expenses. But things like taking your research group
out for a meal can be expensed during gift funding (but not government funding).
41
Indirect costs not being levied can make a huge difference. For example, in our budget from
earlier, if you were using gift funding to fund the student, you don’t have to pay indirect costs,
and thus get over $30K back. Considering a student costs $63K, you get to sponsor the student
for 1.5 years rather than 1 year!

How should you use your funding? Let us say that you are lucky enough to have all four kinds of
funding. And you have an expense that could be assigned to any of these funding (note that this
is often not the case: you can’t use a grant on topic X to support a student working on Y if X and
Y are not closely related). What should you use?

The typical rule is “Use the most restricted funding that will expire the earliest”. Thus, you should
always use the appropriate grant/contract funding first, then your university funding, and finally
gift funding. Gift funding is the most flexible, so use it only when you have no other option.

Always spend your gift funding on expenses that will have overhead (if funded from grants),
such as students. Do not spend it on things like equipment (be careful to ask what counts as
equipment at your university) which do not have overhead even when expensed to grants.

Writing a grant with other PIs. While you could write smaller grants as the sole PI, typically for
larger grants, you need to collaborate with other PIs and submit as a team. For example, you
cannot submit NSF Large grants as a solo PI. One thing to remember though: all the funding is
shared between the PIs. For example, imagine a grant for one million USD. It sounds amazing!
However, if the grant was awarded to a team of two PIs, and the grant duration is five years,
each PI can only support one student on it for the grant duration (100K * 2 * 5). Similarly, if a
grant of 10 million USD was won by a team of five PIs, each PI can only support one student for
four years (100K * 5 * 4). So in the end, the share of each PI is similar or lower than that of an
NSF CAREER.

This is not to say you should not collaborate with other PIs to write grants. Collaboration often
brings new ideas and research directions that are unlikely or impossible when working alone.
And bigger grants often provide prestige: for example, landing an NSF Large grant is significant
and puts you on the map. But it is useful to calculate how much you expect to receive if a grant
proposal is funded.

3.3 The grant application process

So how does one go about submitting a grant? There are several steps in the process:

1. The funding agency first puts out a call for proposals. For example, check out the NSF
call for CRII grants, a startup grant meant for new PIs who have not yet received a grant.
Note that PIs at many R1 universities are not eligible for CRII. This was posted on May 18,
2022. Each call has a solicitation page with details about the program and how the
proposal should be structured – you should read this carefully.

42
 2. In some cases, there are webinars where the program managers explain the grant
opportunity. CRII had such a webinar on Jul 13 2022. I highly recommend attending
these webinars if it is your first time submitting for a particular grant opportunity.

3. In some cases, the funding agency might want you to submit a preliminary proposal.
They will review the preliminary proposals and inform you if they would like you to go
ahead with the full proposal. This is not the case for CRII.

4. There will be a due date by which you need to submit the proposal. For the CRII, this is
Sep 19 2022. The proposal will have a number of parts, and the solicitation will clearly lay
out what the proposal should look like.

5. You write the proposal, and sometime before you submit, you send it to the Office of
Sponsored Research (OSP or OSR) at your university. They have to approve the proposal
before you can submit, and they may request to see your proposal a week or 10 days
before you submit. Please make sure you get a draft to the OSP on time!

6. Finally, you submit the proposal, and you wait. To be exact, your admin/OSP submits for
you, so you want to let them know when you are done. You and your admin/OSP will be
submitting a lot of proposals together; make sure you get to know the folks doing this. I
only have experience with NSF proposals, but these usually take 6-8 months to be
reviewed. In some cases, it might take even longer.

7. If you are lucky enough to get your proposal funded, you might get an email or a call from
the NSF program manager. They might ask you to slightly modify your budget. For
example, I was asked to reduce my NSF CAREER budget (and appropriately reduce the
scope of my tasks as well). You will be asked to provide an abstract that will be displayed
on the NSF website. For example, check out the abstract for my CAREER award.

8. Finally, you will get an official email that your grant has been awarded, it will become
public on the NSF website, and you will get the first installment delivered to your
university. You can then begin spending it in accordance with what you proposed.

9. In the unlucky case that your proposal is not funded, you will get back three or more
reviews from NSF. These reviews tend to be short, and nowhere as detailed as paper
reviews. As such, it can be hard to figure out what to do after getting a rejection. My
advice is to go talk to senior faculty in your department if this happens, and they should
be able to guide you in the right direction.

3.3 Writing your first grant proposal

Writing your first grant can be a stressful experience, since it is quite different from writing a
research paper. There are a number of differences.

First, the audience for a grant proposal is broader than that of a research paper; it is guaranteed
that non-experts in your area will be reading and evaluating your proposal (for example, if I write

43
a grant proposal about storage research, I can expect it to be read by someone working in
systems broadly, but not necessarily someone doing storage research). As a result, it has to be
much broader and more accessible than a research paper. Proposal reviewers might also have a
smaller amount of time to spend on each proposal – some funding agencies ask reviewers to
read a dozen proposals in around a month when you serve on their panel. You can imagine how
much time each proposal would get. Thus, your proposal must be easy to skim and read, must
get to the major points quickly, and must be accessible to a broad audience.

Second, in a research paper, you are reporting on something that is already done. The focus is
on the technical details: how you were able to achieve the result, what are the drawbacks, etc.
However, in a grant proposal, you are talking about things that you will do in the future. In a grant
proposal, the most important thing is to be able to inspire the reviewer, and convince them that
this is a thing worth doing (and funding). In this respect, it is much more similar to a founder
pitching before an investor, than a professor talking to her peers about her research. While the
details do matter to some extent, what is much more important is that you paint a big picture of
how your proposed research will make the world a better place. This is quite different from
writing a research paper, and takes some practice.

For writing your first grant proposal, I’d recommend two things. First, talk to folks who have
gotten the grant you are submitting for, and ask if you could see their grant proposal. Many folks
are happy to share privately. Some folks, like Jeff Bigham and Gillian Hayes, are publicly sharing
some of their proposals. Second, write up your draft proposal, and show it to senior folks in your
department. They would be able to catch errors that could cause rejection.

When submitting grant proposals, do not be disheartened if it gets rejected. As far as I can tell,
most proposals, even from experienced faculty, get rejected a few times before getting
accepted. Grant rejections are a bit more painful because the waiting time is high, you might
only get to submit to a grant opportunity once per year. But with time and repeated submissions,
many grant proposals do get funded.

3.4 Serving on grant panels

Once you become a PI, you will get invited to serve on grant review committees. You can directly
email program managers and say you are interested in reviewing, and they will see if they can
recruit you for the next reviewing cycle. I highly recommend doing this if you can: it is really
eye-opening to see how the sausage gets made! Serving on a grant review committee is
educational: you will immediately pick up what you need to do to make your proposals better.

If you do volunteer for grant proposal reviewing, expect this to take up a lot of time. You will be
asked to review a dozen proposals (or more) in a month or so. It will be similar to a program
committee meeting, where all the proposals will be discussed. You will be expected to share
your view of the paper, and argue for why it should be funded or not.

44
3.5 Getting funding from industry
Since I started at UT, I have been fortunate to receive funding from a number of companies:
Meta, Google, VMware, and Toyota. How did I get this funding? Unlike with government
agencies, there is no set process to follow in general (exception being Google with the Google
Research award).

Giving talks and networking. The process is more organic: I give talks at various companies
about my research, and talk to their engineers about how my group’s research can impact them.
This is useful to gain an understanding of how our work will play out in the real world, and to
understand the problems they are facing.

Sometimes after talks like these, the company reaches out to see if they can sponsor our
research. At other times, the company is looking for academic groups to invest in, and my talk
puts our group on their radar. Note that you might sometimes receive funding even without
doing this – good work is always noticed and rewarded (especially if close to the company’s
interests).

In rare cases, if you know the folks at the company really well, you can also reach out and say I
am trying to achieve research goal X, is your company looking to fund anybody in this space.
But you only do this after building up strong ties with the company over a long period of time.
You can’t start out by saying “can you give me some money”.

Networking and building connections always pay out over the long term. So you shouldn’t have
the mindset of “I gave a talk at company X, I should get some funding from them soon”. Rather,
you should aim to get your group’s research out there and known by companies in that space.
The funding will follow later, but time spent networking and giving talks is never wasted.

3.6 Financial planning for your research group

Finally, I want to talk a little bit about financial planning for your group. As a professor, you are
now a manager, and you should ensure that the students working with you are adequately
funded. If you are unable to fund your graduate students, your department might be able to fund
them as TAs, though the dept would need some advance notice. For postdocs, this option is not
there, so you need to make sure you have funding for your postdocs.

Runway. In general, it is good practice to think six months to 1 year ahead, and make sure you
have enough funding for existing students/postdocs and folks you are planning to hire. Initially,
you will have your startup funding, so you start with a full tank of gas. For every
student/postdoc you hire, you should check how much runway your group has: how long until
your group runs out of funding. You should ensure you have a runway of 6 months to 1 year.

45
This means that if you have 1 year of runway, you need to write grant proposals now (since it
usually takes about a year to be reviewed and funded). At the start of every semester, review
your finances with your admin and make sure you have enough of a runway.

Savings. An important part of financial planning is to always have a little bit extra in your coffers
than your anticipated expenses. For example, imagine that you got a paper accepted, and there
is no associated grant you can expense it to. If you don’t have some money left over, you would
not be able to sponsor your student attending the conference and presenting the paper.

Summers. Part of this financial planning is also deciding what to do in your summers. I strongly
advocate ensuring that you can both pay yourself for the summer, and pay your students. This
might require being conservative in taking on students until you are sure you can pay them.
There is the temptation to cut back on your own summer salary so that you can pay your
students; I would recommend avoiding this. While it can be a temporary fix, it is necessary for
your long-term health, and that of your group, if you maintain your own summer payment.

If you are able to work at a company for the summer, this can reduce the strain on your group’s
finances while also helping your research become more grounded. I spent one summer at
VMware after joining UT Austin, and that time spent was useful for my research. You would
probably need to reach out much earlier to set this up – the group hosting you would need to
make sure they have the budget to host you for the summer.

Coordinating with admins. You want to make sure you and your admin are on the same page
regarding what funding to use for each expense (sometimes it can be hard to move expenses
between accounts). Most admins are experienced and get the picture quickly once you describe
your overall strategy.

Acknowledgements

Thanks to Gabriel Parmer, Indranil Gupta, Jeff Bigham, Michael Ekstrand, Michael Freedman,
Satish Puri, Siddharth Joshi, Solel Pirelli, and Stefan Savage for pointing out errors and providing
suggestions for the chapter.

Summary
In this chapter, we talked about how grants work, what the application process looks like, and
different sources of funding for research. We went over an example budget that shows all the
different overheads involved in funding research.

We have now discussed two of the four main big roles of a professor: managing students and
managing funding. Next, we will discuss teaching.

46
Chapter 4: Teaching
In this chapter, I’m going to talk about one of the main responsibilities for a professor: teaching.
As students, you would have seen the classroom lectures and some part of the grading involved
in teaching. If you were a teaching assistant (TA), you would have seen a lot more of the
grading. But there is a lot more to teaching, as you will soon find out!

Teaching is one of the parts of my job that I enjoy the most, and it is the one of the main
reasons I became a professor. There is nothing like explaining something to a group of
students, and seeing the light bulb go off in their heads. A class filled with bright, motivated
students, one that is interactive and full of back-and-forth, is a joy to teach.

Another benefit you get as the teacher is understanding the material deeply – students are
going to ask you questions you may not have thought about yourself, questions that make you
look at the material in a new way. Sometimes, questions from students spur new research!
Many professors have spoken about the beneficial effect of teaching on their research.

4.1 Teaching at an R1 university

Overall, teaching an undergraduate or graduate course at an R1 university in the USA involves
four things: designing the course (including exams, projects, and assignments), classroom
lectures, grading, and course administration (including managing TAs and handling academic
dishonesty). I’ll cover each of these in detail in later subsections.

The teaching load (number of courses you teach per year) varies from two to four at R1
universities. It is higher at more teaching-focussed universities where you might be teaching
two or more courses each semester/quarter. Typically, the department will contact you at least
one semester before (sometimes one year before) to figure out what you will be teaching. For
example, you will be asked in Fall 2023 what you would like to teach in Spring 23. Sometimes,
you are asked in Fall 23 what you would like to teach in Fall 24! So it is good to plan ahead.

Balancing teaching and other responsibilities

Note that the importance of teaching varies depending on the type of institution you are at. If
you are at an undergrad-focused university, this is an important part of your job, and it will be a
significant part of your evaluation. At such universities, you will be teaching multiple courses
each semester, and it is not uncommon to have teaching loads of four courses or more per year.

At an R1 research university, the most important thing you are judged on is research; you can
get tenure doing great research and being a mediocre teacher, but you cannot get tenure doing
mediocre research and being a great teacher. Of course, if you are a terrible teacher, you might
not get tenure (every part of your tenure application carries non-zero weight). One potential

47
failure mode for new profs is to focus on teaching to the neglect of their research: make sure
you don’t do this.

This is not to say you should do a bad job at teaching or that you shouldn’t spend time on it. But
you can always spend more time on teaching – you can always make the slides better, spend
time on demos, etc. So it is important to carefully manage how much time you spend on
teaching.

When I started as a professor, I used to spend about 2 days a week on teaching. I had “teaching”
days each week where I knew I wouldn’t get anything else done. I would focus on research and
service on the other days of the week, and make sure my teaching responsibilities didn’t
overflow too much (sometimes it will, like during midterms etc). I would not recommend
spending more than 3 days a week on teaching-related activities.

How is teaching evaluated?

So how will you know when you have done a good job? How will you be evaluated?

Every semester, students will be asked to take a survey about your course. This will happen
before the final exams, during the final days of the class. Students will be asked to answer
questions like “Is the course well-organized?”, “Was the instructor available for office hours?”
and whether they would recommend the course to others. These surveys are part of how the
department evaluates you – they will look at the survey scores, and trend of the scores. Are you
improving at teaching over time? Are you fixing problems you faced in one instance of the
course in the next instance?

These surveys do have some problems, though. For example, female instructors and instructors
from under-represented minorities often get worse scores simply due to their identity (see this
and this). It is known that scores are affected by things like whether you give your students
cookies. This is all known by the administrative folks as well, so they take a nuanced view of
your teaching evaluation scores.

As part of the tenure process, your department might also have one or two faculty members sit
in your class and observe your class. They will then provide feedback, both to you and the
department. This is useful since it is a different kind of feedback to the ones the students give
you, more considered and grounded in pedagogy.

Sometimes, students will just email at the end of the semester to tell you they enjoyed the
course. These emails are the best, and you should treasure them. It will be handy to go back in
rough times and look at these emails, to remind yourself why you do what you do.

Planning your teaching

Teaching your first course. At some R1 research universities, you get one or two semesters of
“teaching relief” as part of your start-up package. This means that you can opt out of teaching
48
for one or two semesters. One way to use this benefit is to not teach your first year, completely
focussing on getting your research lab set up. I would advise against doing so.

Teaching in your first semester has a number of benefits. First, teaching a graduate-level course
in your area of interest is an excellent way to find your first students (it is how I found my first
students). Second, it is useful to front-load the work of preparing for a class – since you are
already on (or should be on) a break from research following your graduation, this is the best
time to prepare for teaching; you don’t want to take time out once your research is going well.
Finally, teaching introduces some structure and regularity to your day – if you are not teaching
and you don’t have any students, it is harder to establish a regular routine at your new
department.

For your first course, I would recommend teaching a graduate-level seminar type course where
the students read and discuss papers, and do research projects. This will attract students who
like your area of interest, and you will be able to evaluate their research potential through the
class. At the end of class, you should be able to tell if some of the best students in class are a
good match for your research group.

Teaching relief. As to when to utilize the teaching relief (if it is offered), I would recommend
spreading it out over a couple of years. When I started at UT, the annual SOSP/OSDI deadline
was in Spring. So I would teach in the fall and focus on research in the spring. Work out a similar
rhythm for your research group.

Teaching multiple sections. At some universities, you can opt to teach multiple sections of a
course in the same semester. This would then count as teaching two separate courses. I know
professors who do this so that they finish off their teaching requirements in one semester, and
then travel in the second semester or focus on research. This can be a viable strategy since it
amortizes some of the teaching effort (such as class preparation and setting exams/projects)
across the different sections; some work like grading cannot be amortized though. This can be
a good option to focus your teaching efforts on one semester.

Teaching a new course. Teaching a course you have already taught is significantly simpler than
first designing and creating a course. Ideally, you would not have to create a course from
scratch: you can borrow the teaching materials of professors at the university who taught the
course before you, and customize it according to your teaching style. I created a new
undergraduate course (Virtualization) in my second year at UT Austin: it was a lot of work, but I
enjoyed doing it. However, I still would not advocate doing this early on: if you are not careful, it
could end up consuming all your time, impacting your research.

4.2 Designing a course

Typically, when you start out as an assistant professor, what you do is take over an existing
course from a faculty member. This is usually the graduate-level course in your area of
expertise. For example, I taught the graduate-level Advanced Operating Systems course when I
49
first started at UT. You should talk to the person who taught it previously, get their slides, notes
and other teaching material (this is common and encouraged, so don’t feel shy). You can then
customize it as required. For the Advanced OS course, Emmett Witchel generously shared his
teaching materials with me, and I built upon it.

Sometimes you might want to design a new graduate-level course, focusing on your expertise.
This can be a great way to find students interested in your area, and train them on the basics
that they would need for your area. For a course based on reading papers and doing research
projects, it will not take up a lot of your time since these are papers you are closely familiar with.

I would definitely not recommend teaching a new undergraduate course as your first course. I
designed a new undergrad course in my 2nd year at UT, and it was a big time investment (more
than I had planned). I would recommend putting this off until you are somewhat established in
your research and then doing it.

When designing a new course (or customizing an old course), it is important to think about the
learning objectives that you want the students to achieve, and then design the course
components to aid this. For example, you might want the students to be familiar with a certain
tool or a certain technique; creating an assignment around this is a good idea. Similarly, you
might want to think about what other concepts they need to understand to do the assignment,
and work backwards to populate the course.

Designing a graduate course

In general, teaching a graduate course is easier than teaching an undergraduate course (though
not as fun, in my opinion). This is because the class sizes for graduate courses are smaller
(anywhere from a few students to 40).

There are two flavors of graduate courses: a reading-based course where the class reads and
discusses research papers (along with doing one or more research projects), and a seminar
course where the professor lectures in each class (in addition to the research project). The
reading-based course, especially if it is your area of interest, needs little preparation, as these
are mostly papers you are familiar with in your research. A certain amount of thought needs to
go into research projects that can be done by junior grad students in about two months. But
overall, a reading-based graduate course is one of the easiest to teach (that’s partly why I
recommend teaching it as your first course).

The second type of graduate course is a seminar course. This is much closer to an
undergraduate course in that you need to prepare for each lecture. This type of course is useful
when undergraduates are also taking the course, since they are more used to this mode of
instruction rather than reading papers. Seminar courses have exams and much more defined
projects (what I would term an “implementation project” rather than a “research project”).

50
A variant of the seminar course is where students present some material (sometimes a
research paper) each class, with the instructor adding context and extra knowledge, and
provoking discussion among students. From the professor’s viewpoint, this is similar to the
reading-based course, and doesn’t take much effort to teach if the subject is in your area of
expertise.

Designing an undergraduate course

An undergraduate course is different from a graduate course in many ways. The biggest
difference is structure – you might tell your graduate students, “let us sit and ponder the
mysteries of the universe (or of distributed systems)” and there is a reasonable chance the grad
students will take you up on that. Undergraduates have no time for this; they are taking 4
courses (or more), and they want classes where it is clear what they need to do to pass/do well
in the course. So your undergraduate course needs to be much more structured.

There are a few more differences. The class size is significantly bigger than a graduate course –
popular undergraduate classes can have hundreds of students! Due to the big class size,
evaluating students on projects and exams becomes a challenge – you need to structure the
class so that you can efficiently (and fairly) evaluate the student submissions. Designing the
class so that TAs can evaluate the student submissions also becomes a priority. For example,
open-ended questions that can be interpreted multiple ways with many possible answers is not
a great fit for an undergraduate class of 100 students (you can have maybe one such question
on the exam, but if all the questions are like this, you are going to have a hard time).

Designing an undergraduate course from scratch consumes a lot of time, both for preparing
lectures and designing projects and exams. It is also a lot of fun, so you have to be careful to
manage the amount of time you pour into your undergraduate course. When designing the
course, it helps to first write down the learning objectives you want students to achieve, and
then design course components that help achieve each learning objective.

Undergraduate students are much more likely to ask for re-grades and extra points than
graduate students (understandable given how much GPA matters for undergraduate students).
To reduce this overhead, it helps to be concrete about each exam question/assignment and
what exactly will fetch points. The clearer students are about what gets points, the fewer
regrade questions you and your TAs will get.

I love teaching undergraduate courses, especially because I teach an upper-level elective

undergraduate course (so all the students in class want to be there, and aren’t taking it to satisfy
a requirement). The undergraduate students at UT Austin are amazing, and I learn something
new about the material everytime I teach due to their questions. The energy in class is amazing,
and it is always interactive. I highly recommend teaching an upper-level undergraduate class in
your area of expertise if it is possible.

51
4.3 Classroom lectures
Classroom lectures are the most visible part of teaching, and (at least to me) the most fun part
about teaching. When you have an interactive class that is buzzing with questions, it is a joy to
teach!

What does good teaching look like?

Good teaching comes in various forms, but the root of it is engaging students. You want
students to listen to your lectures, to think and understand the material, and absorb the main
points that you are trying to communicate.

I’ve found that many good teachers are enthusiastic about whatever it is teaching, and pass on
that enthusiasm to their students. It is hard to be a good teacher if you are not interested in the
material – this might happen if your department is short on staff and needs you to teach a
course you would rather not teach.

Given these two things, engaging students and being enthusiastic, the other mechanics of how
you teach can vary a lot. For example, I like to teach without slides, using the blackboard – a
style I picked up from great teachers such as Remzi Arpaci-Dusseau, Michael Swift, and Jeff
Naughton at Wisconsin. I find that my students are more engaged when there isn’t a projected
slide in front of them. I share notes with students so that they can review the material later. The
act of drawing figures on the blackboard also slows me down, and gives students time to think
and ask questions.

Some of my colleagues at UT Austin like to teach with slides, and are immensely successful at
it. So I don’t think it is the media that matters, it is about how you engage your students.

Other tips for lectures

Some basic tips I’ve gotten over the years:
- Ask questions in class, and give students 30 seconds to a minute to respond. When you
first do this, the wait can get uncomfortable for you (and for your students). But do it
anyway – maybe drink some water while waiting for an answer. If your students see that
you are serious about questions, they will ask.
- Students learn how the dynamics of the course work from the first few classes. If in the
first few classes you are not asking questions (and establishing that culture), it will be
harder to do it later on.
- A common mistake is to try and cover too much in each lecture. Despite your best
efforts, students are not going to absorb or remember everything. Think about what the
major points you want to communicate are, and focus the lecture around those.
- Start each class with a recap of what was covered in the previous class, and finish each
class with a summary of what you had just talked about
- Something I like to do: if your class is scheduled for 75 minutes, wrap up the lecture at
around 65 minutes. Give students about 10 minutes to get to the next class, and to come
52
 up and ask questions. Many students are shy and won’t raise their hands, but will come
and ask questions at the end of the class.
- Different students excel at different things. Some do well on exams. Some do well on
projects. If possible, design your course to offer a way for all these students to shine.
- If you see that a student is doing well (either in class, or when grading their exams), add a
note saying so, and that you think they might do well in grad school. It might be the first
time they ever think about that career path. I’ve successfully nudged several students
towards grad school in this manner.
- It is inevitable that some students will not like your class. Part of it would be what they
are going through themselves – undergrads lead incredibly busy lives. So don’t take it
personally.
- Similarly, it is inevitable that there will be students in your class who just want to pass
and get this over with (this happens more if students are required to take your class, less
so in elective classes). Don’t take this personally either. There will be others who will take
your class out of interest, focus on them.

4.4 Grading
Grading exams and assignments is one of the least enjoyable aspects of teaching. Grading and
course design go hand in hand – how the course is designed affects how easy or fair it is to
grade.

When grading, we want to achieve two goals. The first to grade fairly. Similar answers or similar
code should get the same points. The points awarded should not depend upon the person
grading. Even with the same person, bias can creep in and other external factors (such as
whether they were hungry when grading) can affect grading.

The second is to grade efficiently. You want to design your assignments and exams such that
you or your TAs can grade each item in a few minutes. The answer for each assignment should
be concrete: you should have the breakdown of how points are going to be awarded (for
example, “2 points if they say X”) or taken off. If it takes hours to grade each student
submission, you are probably doing something wrong. While the grading time definitely goes up
as the number of students increases, you can manage this by automating the grading as much
as possible. For example, you can have a grading script that you run on student submissions
that automatically determines the grade. One potential problem with automating grading is it is
harder to discover students cheating; you might want to have your TAs scan student
submissions anyway to catch obvious cheating.

Grading order. When grading exams, do not grade all the questions for each student: if the
student does well on an earlier question, it introduces bias when grading the later questions;
instead, grade one question for all students in class and then grade the next question. Have the
student write their name and ID *only* on the first sheet, such that it is not known when you are
grading the other questions. When distributing the grading work among the TAs, do not split

53
based on students – it will lead to uneven grading; instead, have each TA grade a set of
questions consistently across the whole class.

4.5 Course Administration

Managing TAs
An important aspect of teaching courses with a lot of students is managing TAs. A rough rule of
thumb is that you want about 30 students for each TA (this number might vary based on the
course); so if you have 90 students in your class, you need three TAs. As the number of TAs
increases, managing them becomes a task in itself. In such circumstances, you should appoint
a “Head TA”: someone who is responsible for making sure all the TAs are on the same page,
handling things like dividing up the work among TAs and scheduling discussion or office hours.

Good TAs are worth their weight in gold. If you can get a good TA to TA for your course again, it
is great: the TA knows the material, and knows what question students will ask. It is usually
less work for them as well, compared to TAing a new course. Always try to have good TAs
return to your course.

Managing your TA’s workload. You want to make sure you aren’t overloading your TA. For each
task you assign the TA, have a rough estimate of how long it would take you to do it; double that
estimate to get a sense of how long it might take the TA. Make sure the total amount of time the
TA requires is not over 20 hours a week (or whatever their assigned work hours are). Holding
office hours and grading are all part of their official workload.

Another part of managing your TA is figuring out what tasks you can offload to them. For
example, “design a project for this course” is probably too high-level for your TA. You could
instead say “here is the project I’ve designed, could you take a shot at it and give me your
feedback”. Make your TA tasks as concrete as possible. With a more senior and experienced TA,
you could give them more freedom in designing assignments. Make sure you guide them on
how to design assignments such that it is easy to grade fairly and efficiently.

Handling academic dishonesty

One of the unfortunate aspects about teaching is that you have to handle situations where
students are cheating in class. This can happen in a variety of ways: you might be grading an
assignment and notice that two submissions are similar, or notice that submitted code looks
identical (sometimes down to comments!). Sometimes the submitted code reads very
differently from the rest of the submission, indicating it has been copied from somewhere.

Why is academic dishonesty bad? One way to look at this is that the person being harmed the
most is the student themself: they are paying a lot of money to get an education, and instead
they are just getting a degree. The things they didn’t learn will become apparent once they
graduate and get a job. However, if cheating is rampant in a course or a university (and publicly

54
known to be so), it can harm the value of the education for all students: the value of the degree
drops significantly even for honest students. Moreover, if the class is graded on a curve (a bad
idea that I don’t recommend), cheating influences the grade of all the students in the class.

So given this, how much effort should you spend on detecting academic dishonesty? You could
run tools such as MOSS to detect similarity between submissions. There are also plagiarism
detectors for text submissions. Canvas (the course management system used at UT) now
provides some of these tools built in, making it easy to run. You can imagine writing your own
tools to detect cheating. There is a big spectrum of how much effort to put into this.

Personally, I use built in/easy-to-use tools for this purpose. I don’t spend a lot of time on
detecting academic dishonesty. My classes use absolute grading (for example, if you score 87,
you are guaranteed to get an B+, regardless of how anyone else in class performs), except for
the top grade (A). This limits the bad impact of cheaters on the rest of the class.

Once you detect cheating, what should you do? Your university might have a protocol for you to
follow. Typically, the first step is to talk to the student. Talk about the rules and expectations for
the course, and the evidence you have for cheating. Sometimes, the students immediately
accept their mistake – they will explain they did it in a moment of panic or fear (or that they
didn’t know what they did was against the rules), and that they are sorry. This is the best
outcome. You explain the penalty to the student (typically, if they cheated on an assignment,
they get zero points for that assignment; if the penalty is too light, it will encourage further
cheating) and implement the penalty. In some cases, the incident and the resolution (whether
the student accepted, and what was the penalty) will have to be reported to Student Affairs.

In other, more unfortunate, cases, the student will not accept that they did anything wrong. In
such cases, you will once again report the incident to Student Affairs, and they will investigate
whether any academic misconduct occurred. They will talk to the student, the TA involved, to
you, look at the evidence, and finally come to a conclusion. This is a big process: depending on
the university, it can be quick (takes about 2 weeks) or extremely long (several months).
Generally, it takes up a lot of everyone’s time, so it is always unpleasant to handle. But once
Student Affairs reaches a decision, the student and the professor are notified, and the
appropriate penalty is then applied.

Due to how expensive (in terms of time) it is to handle misconduct cases, it is worth trying to
design your course and your assignments and exams to minimize opportunities for misconduct.
For example, if your exam is online, and your exam questions are google-able, a small
percentage of students will cheat. I find proctoring software repulsive and degrading, with
significant impact to the dignity and privacy of those who take exams online. So the solution
turns out to be to design exam questions that require application of material learnt in class,
such that it is hard to google the answer in a couple of hours (as a side note, this also makes
the exam much more fun!). This requires more effort on your part in setting exams and
assignments (especially coding projects, where it is not so easy to come up and test new

55
projects) – so the tradeoff is between how much time you spend creating exams and
assignments versus how much time you spend in meetings with Student Affairs committees.

Handling requests for extensions

Another administrative aspect of teaching is handling requests from students for extensions on
their assignments. Students often email and ask for extensions, with reasons ranging from “I
was sick” to “my parent/grandparent/significant other is sick”.

There are two things to consider here. First, do you require documentation to prove that their
reason for the extension is valid? For example, you could ask for a doctor’s note if they say they
are sick. Second, assuming the reason is valid, do you grant the extension? For what reasons do
you grant extensions? For example, do you grant it if the student is personally sick? What if their
grandparent died?

At some universities, if you are a student and you are sick (or have an emergency), you can
contact a service similar to UT’s Student Emergency Services . They will verify your
sickness/emergency, and then contact professors for all your courses to get extra time on your
behalf. This is great when it works! But I find the % of students who contact Student Emergency
Services when sick/having an emergency to be low.

Here is how I handle requests for extensions. Typically, I do not ask for documentation for
sickness or other reasons for an extension. I cannot imagine saying, “please prove that your
grandparent died”. I take the student’s requests at face value and assume they are speaking the
truth. I generally grant extensions if the reason was something out of their control – sickness on
their part or family, internet outages, literal war (for students of my online masters course in
Ukraine).

Why am I not inundated with fake requests for extensions? I always include grace/late days in
my course that each student can use for any reason on assignments (if you submit a day late
and use one grace day, it will be as if you submitted on time). I am generous with the grace days
– typically there are 10 grace days each semester. I find that if a student is contacting me
beyond using the grace days for an extension, it is something serious.

Overall, I think that there is a trade-off between being merciful and being fair. If you give
someone an extension because of some circumstances in their life, it is not fair for all the other
students who did not contact you, but nevertheless had troubles. The students are not really
completing the assignment under the same constraints, so you could argue that to be fair, you
must decline all extensions unless health-related. I personally find that being merciful, and
helping students in trouble who reach out to you, results in a much better course experience
overall (both for students and for you).

56
Summary
Teaching is an important, and rewarding, part of your role as a professor. We discussed
teaching at an R1 university, designing a course, and ways to make your lectures engaging. We
also discussed how to grade fairly, and administration of your course.

Next, we will chat about doing research as a professor. I discuss this last among the main job
responsibilities, as this is something grad students are directly trained for before starting as a
professor.

57
Chapter 5: Research
Now let us talk about the central aspect of being a professor at an R1 research university: doing
research. This is in some sense the easiest part of the job, because this is what you get directly
trained for as a graduate student; in another sense, this is the hardest part of the job, filled with
uncertainty, ambiguity, and rejection. Your success as a professor at an R1 university will
depend in large part on whether you can come up with and execute a strong research vision.

We have already talked about getting funding and finding students, so I will focus on doing
research once you have funding and students.

5.1 Evaluating research from the point of view of tenure

Let us first talk about what the research component in a successful tenure case will look like. It
is always good to know what the end goal is before we start! Success is not defined purely by
whether you get tenure, but since all tenure-track professors wish to get tenure, it is useful to
know the requirements.

Abstractly, the tenure committee is trying to determine if you did good research. Concretely, the
tenure committee is concerned with two things: the publication record, and letters. The
committee looks at whether you have been consistently publishing at the top venues for your
area of research. The committee will request letters from the experts in your area of research;
these letters should make the case for granting or not granting you tenure.

What makes a publication record good? One thing the committee will look for is consistency: if
you publish a lot in year one, but nothing for several years after that, that will result in some
questions. Another thing that will be examined is productivity, relative to your field. For
example, systems projects take about a year to complete, so the committee might look for one
or two strong publications each year for systems professors. The rate of publication is much
higher in areas like machine learning, so the bar will be different for that area. You should talk to
senior folks in your department to get a sense of what they are looking for.

What do good letters look like? At a high level, your letters make the case for granting you
tenure. They talk about your body of work, some of your main results, and your impact on the
field. One important thing to note here is that the letters never talk about more than three papers
or projects. Regardless of how many papers you publish, your letter writers will focus on your
most famous/most impactful work. Thus, it is important that there are a few papers or a
coherent line of work that you are well-known for; simply publishing a lot of papers is not
enough for tenure. When folks think of you, you want them to remember a line of work. "John
Doe is well known for introducing X as a way to solve Y".

58
Note that this places constraints on the kind of research you can do. If your line of work is
obscure and not well known to the community, you will not get good letters. It also has
implications for outreach; it doesn't matter how good your work is, if nobody hears about it (and
writes about it in your letters). Thus, you must do both high-quality work and ensure that the
community knows about it.

Another important thing that the committee is trying to ascertain is your research
independence. Prior to becoming an assistant professor, most (if not all) of your papers would
be co-authored with your advisor. It is important that you show that you can lead research
projects on your own after becoming an assistant professor; if all your papers after becoming
an assistant professor are still with your co-author, it raises red flags. Similarly, it is good for
your research portfolio to contain at least a few projects that are just you and your students.

5.2 Coming up with a research agenda

What is a research agenda? How is it different from a research project? A research agenda
sets a broad direction for multiple research projects. You can think of it as a general direction
towards a broad goal. For example, the broad theme/agenda of my NSF CAREER was "building
IO-efficient infrastructure". Your research projects are then concrete tasks that advance this
broad goal; for example, one of the projects under my CAREER agenda was building a
write-optimized key-value store.

Coming up with a compelling research agenda is one of the hardest things you will do as an
assistant professor. By the end of your PhD, you will have some inklings of what you would like
to work on in the future. However, translating that into a concrete research agenda that can
inspire students and funding agencies is quite challenging. In some sense, solving this
challenge is how you write your first grant proposal (such as the NSF CAREER). All grant
proposals require a compelling, multi-year research agenda.

My advice for coming up with a research agenda would be: pick the intersection of inspiring,
technically challenging, and doable. I'll talk about inspiring first, because I think this is where I
and many other new profs fail. Simply put, your research has to inspire someone who is not an
expert in your sub-area. Why is this the case? Because your research agenda is typically
important for things like grant proposals or awards, where you are evaluated by peers broadly in
your field, but not necessarily in your sub-area. This is in contrast to paper submissions, which
are evaluated by area experts. For example, in my case, my research should inspire folks in the
broad systems community, not just storage. One of the mistakes I did early on was send in a
narrow proposal that was technically challenging and doable, but not inspiring; it got rejected.
You should both be able to explain what your intended research is to a lay person, and get them
to see why it is important.

The second important criteria is picking something that is technically challenging. Without a
hard technical problem to solve, you will have trouble in both getting funding agencies to fund
your research, and in getting your papers published in your community. So you need to think
59
about your broad goal and identify the technically challenging aspects. These aspects need to
be something that cannot be solved with the current state-of-the-art.

Finally, you need to pick a research agenda where you bring something to the table. Funding
agencies like to fund work where the expertise of the PI is essential for the work. This is also the
niche where you are likely to make fast progress, given your skills. So I would definitely
recommend having your first proposal be something tied to your proven track record, not
something tied to what you want to do.

This can be a little tricky if you are trying to switch sub-areas. Perhaps you worked on topic X
during your PhD, and you are sick of it, and want to move to topic Y. You now have a bit of a
bootstrapping problem, in that funding agencies won't fund you for Y unless you have
demonstrated expertise. Thankfully, many R1 schools provide startup packages that help
bootstrap research; you can use this to publish some research on topic Y, build up your
credibility, and then apply for a grant proposal on Y.

5.3 Picking research problems

I think problem selection is one of the most important skills as a researcher, especially because
we have so much freedom in deciding what to work on. The best researchers hone in on
important problems and focus their limited time and attention on these problems.

An easy way to build a poor research portfolio is to simply pursue every idea that you come
across. You might imagine that great researchers simply just have one great idea after another.
However, from working with great collaborators, I've learned this isn't the case. These
researchers get OKish or even bad ideas just like anybody else; the key difference is that they
reject potentially publishable but OKish ideas until they get to the really good ideas. So you
should learn to say no. You have limited time and energy, as do your students; typically, each
student is only going to work on 3 to 4 papers before they graduate.You should make sure those
papers are good ideas, worthy of the many months or years of time that are collectively spent
on each idea.

So how do you pick research problems to work on? One of my professors at Wisconsin used to
say, "It should be beautiful, or it should be useful". I think that captures it concisely: the criteria
you are looking for are elegance (or how clever or neat an idea) is, or utility (it should be of use
practically). Arun Kumar at UCSD says something similar in his excellent blog post on the topic.
NSF also has two key goals for each proposal: intellectual merit (how does it advance the
state-of-the-art) and broader impacts (how does it positively affect society).

Research ideas you consider will have some mix of these two qualities. At one extreme, an idea
might be impractical, but open up new lines of thought. At the other extreme, an idea might be
useful, improving the state of the practice significantly. Both these extremes have problems:
working on ideas without considering utility may lead to ivory-tower impractical theory only

60
useful for other academics;working on practical ideas without regard for advancing the state of
the art transforms your lab into a poorly funded industrial research and development center.

Since I work on systems research, I tend to skew towards the utility side; I like building
prototypes that can be used by real people. But I avoid ideas where there isn't an interesting idea
behind the prototype, and it is clear from the start what needs to be done to solve the problem. I
think the sweet spot is in the middle, where there are both intellectual challenges and practical
utility to users.

5.4 Knowing when to give up

An important part of managing your research is knowing when to give up or abandon a project.
This might happen due to various reasons: the project might be too hard, it may not be a good
match for your skills or your student's skills, or it may take more time than you want to spend on
it (the student might be graduating). In any case, it is important to identify and handle such
situations correctly.

Projects have a life of their own -- once you have spent more than 6 months on a project,
everyone involved will feel reluctant to pull the plug. "We have already spent so much time on
this, might as well see it through to the end" -- everything ends up thinking along these lines, but
it may not necessarily be the right thing to do.

A good way to prevent this is to be extra careful when starting a project. Always try to ascertain
whether your team has the skills required for the project. For example, starting a theory-heavy
project with a student who has amazing systems-building skills is not a good fit. Before starting
the project, discuss the "failure scenario", and come to an agreement: "If we haven't made good
progress towards X in three months, we will pivot/try something else". Now, "making good
progress" is still very fuzzy, but it can be hard to come up with clear-cut objective criteria. At the
end of the period, you and your team should make the call on whether to move forward.

My recommendation is to cull early, and review often. If a project doesn't look exciting or there is
little progress (or even hope of progress), switch to something else. Research is interesting in
that (to borrow a sports metaphor from Warren Buffet (!)), you don't have to swing at every ball.
Wait for the fat pitch, where you have high confidence going in that you can make progress.

5.5 Managing collaborations

Another important aspect of managing your research is managing your collaborations. A good
way to spread yourself too thin is to say yes to too many collaborations. The social aspect of
collaborations makes it hard to say yes, especially if you respect and admire the folks asking
yes.

But you should remember your time and energy are limited. Ask yourself whether the
collaboration is right for you: does it extend or apply some of your work, in a domain you care
61
about? Is there a chance for interesting ideas (in your sub-area) to come out of the project? Will
it help in practical adoption of your ideas? Does it help drive the broad goals of your research
agenda?

I think a couple of fun collaborations that are not related to your research are okay. But if you
spread yourself too thin, you risk losing momentum and energy for your main line of research,
which is what you hope to be known for anyway. So I would recommend being careful with
starting new collaborations.

Industrial collaborations
Industrial collaborations are different from academic collaborations in many ways. Let us first
talk about a collaboration where the company is sponsoring research: they are paying you
and/or your student to conduct research. Doing something like this will require a Sponsored
Research Agreement. Typically, the Office of Sponsored Research will be involved (and they
might involve some lawyers). The Agreement will spell out how the Intellectual Property (IP)
generated during the research will be shared, and how patents arising out of the research will be
handled. Sometimes, there are templates for these things, everyone agrees and it goes
smoothly. At other times, it can involve a lot of back and forth before the agreement is in place.

A second type of industrial collaboration is where you are just collaborating with researchers at
a company. The company is not paying you or your students in this case. By default, the
intellectual property you and your students generate will belong to the university (though some
universities have exceptions where the IP belongs to the professor). In this setup, it becomes
tricky if any shared patents are to be filed (this is where having an agreement would help).
However, in the absence of any patent filing, it is similar to collaborating with other academic
researchers.

Regardless of the type of industrial collaboration, some companies have extra steps to be
followed before their researchers can publish. Sometime before submission, industrial
researchers need to submit the draft to an internal committee who will read and ensure that no
company secrets are being revealed. In some cases, the committee will provide technical
feedback that can help strengthen the paper. In the vast majority of cases, these checks are
routine and the committee will pass the paper; occasionally though, the company will prevent its
employees from co-authoring the paper unless changes are made.

62
Summary
We discussed how to come up with a research agenda, one of the central tasks of a professor.
We also talked about how to pick research problems, and how to decide to give up on a research
angle. Finally, we talked about how to manage collaborations – a crucial skill for professors.

We have now covered all the central responsibilities of a professor at an R1 university. The next
few chapters will talk about doing outreach, service for your community, and taking care of
yourself while doing this job.

63
Chapter 6: Outreach
Being a professor is like being the CEO of a small company. Just as a CEO appears at various
public events, talks to the media, gives interviews, etc. to promote his company, it is the duty of
a professor to promote the work of her group. Similar to the adage about the tree falling in the
forest, it doesn't matter how good your group's work is, if no one knows about the work.

There is this wrong notion that “good work speaks for itself”. This is a highly damaging myth
because it makes students think all they have to do is to do the research, and the rest will take
care of itself. This is not true at all!

If you observe any star researcher, you’ll notice two things. First, they are mainly known for a few
key results. Nobody remembers all the work someone else did; we tend to remember the best
work. Second, if you go look at their travel, they would have spent a considerable amount of
effort in spreading the word about their key results. They would have written blog posts, given
talks at various universities, and talked about their work on social media.

I really like this model of work: you work on a few ideas, but once you have spent multiple years
on a paper, you then spend weeks or months spreading the word about your ideas. Given how
much time we spend on research, I think we owe it to ourselves to spend some time on
outreach about the work.

Contrast this with a model where you publish a lot, but not a lot of people know about your new
ideas. Though assistant professors do need to publish and be productive to get tenure, the end
goal is to have impact through your ideas, and outreach is crucial to achieving this goal. So let
us talk about how to do outreach.

6.1 Web page

It is kind of strange to begin a chapter on outreach by talking about your web page, but I am
doing so because it is so basic and fundamental. Your university web page should have the
following for every paper you publish:
● Publication metadata like title, authors, and venue
● A link to the paper PDF
● A link to a self-hosted, well-formatted bibtex entry (don’t depend on ACM for this)
● A link to the talk slides if the work has been presented
● A link to the Github repository if the code is available

I find it baffling how many professors don’t put links to their paper PDF in their web pages. Many
publishers allow authors to post the PDF (the “camera-ready” accepted version, minus copy

64
editing and formatting applied by the publisher) on their own web pages; for example, see
ACM’s Author Rights page and IEEE’s Author Rights page. With open-access publishers like
USENIX, it is explicitly allowed to post the published version on your web page. It is much easier
to google a paper and pick it up from the author’s web page rather than from ACM or IEEE. You
should make it as easy as possible for others to find and read your paper: please share it on
your web page.

You should also make it easy for people to cite your work. Writing your own bibtex entry gives
you a chance to get the details right, such as capitalization of the title in your paper, or perhaps
the spelling or accents in your co-authors names. Naming the bibtex entry right also saves the
folks who cite your paper some work: it is easier to cite “MohanEtAl21-DataStalls” rather than
“10.14778/3446095.3446100”.

Finally, I believe that sharing your code so that others can reproduce and build on your work not
only leads to better science, but also increases the impact of your work. I highly recommend
sharing your code whenever possible. Add a link to the github repo on your web page (and
inside the paper) so that people can find your code easily.

6.2 Blogging
Blogging is one of the best ways to get your ideas out. Setting up a blog is free and quick, and
writing is a skill that professors are already good at. Some professors have blogs that are read
by thousands (or even tens of thousands) of people – for example, see the widely-read blog,
“Shtetl-Optimized”, by my colleague Scott Aaronson.

There are several benefits to blogging:

● By blogging about your area of expertise, you can quickly build an audience of informed
readers who engage with your work.
● A blog is a great way to create your “brand” – the topics you know and care about (so
that someone could invite you for a panel or a PC).
● A blog is a good way to reach out to a broader community, increasing the impact of your
work. For example, practitioners in industry might not read a technical paper, but will (and
do) read blogs
● A good blog is also an effective recruiting device: students will read it, get inspired, and
want to work with you.

What makes a blog good? A good blog is accessible, interesting, and regular. First and foremost,
a blog should be accessible – a blog that simply regurgitates the technical material in your
papers is not going to be widely read. The audience for a research submission is different from
that of a blog: for a paper submission, you are mainly talking to your peers in the academic
community, trying to get them to accept your paper; for a blog post, you are talking to a wider
audience: interested practitioners in industry, undergraduate students, etc. So you don’t need to
be as defensive, or talk as much about the novelty or how it compares to prior work. You should
write it so that an interested third-year undergraduate student can read and appreciate the
65
material. This means no unnecessary use of acronyms or technical terms, and no assumption
that the reader would have read the seminal papers in the field. You should focus on the big
picture: the big problem that the paper tackles, what is the cool new insight that is powering the
paper, what are limitations (practitioners care about where this can be applied), and what are
the implications for the wider field and folks in industry.

Second, a blog must be interesting. Oftentimes, this comes directly from the research that you
are doing. When we write a paper, we tell the “straight-line story”: we had this brilliant idea, we
implemented it, and we got amazing results. However, as researchers, we know that the path of
research is almost never a straight line; there are many twists and turns and dead-ends.
However, a research paper is not the right place to tell that particular story. A blog post is
appropriate, and hearing such stories is inspirational and educational to the junior members of
our research community. For example, it is a big deal for a junior graduate student to
understand that even their research heroes get their papers rejected sometimes! A while ago, I
started a blog called “Computer Science Research Tales” to document these stories behind CS
research. Such tales would make excellent blog posts!

Finally, the hardest part about blogging is being regular about it. For a blog to really take off, it
needs to have regular updates. This is pretty hard to do! For example, when I started at UT, I
used to blog regularly about my research, but it was harder to make the time after we had kids.
Ideally, you want to have a post every month or so. The most popular blogs often have multiple
posts a month!

How to get started with blogging. I would recommend just making it part of the publication
process. Basically, every time you have a paper out, spend some time writing up a blog post
about it. The material would be fresh in your head, so it would be easy to write the blog post and
talk about the path the paper took to getting published. When you post the details of the paper
on your web page, link to the blog post as well. When someone asks what your latest paper is
about, send them both the blog post link and the paper link. I guarantee they will read the blog
post first (and perhaps only the blog post).

6.3 Using social media

I often get asked about using social media as a professor. There are many questions: should
profs use social media in the first place? Will posting on social media make professors seem as
if they are not serious? What should they post? How much time should they spend on this?

Social media is an amazing tool that was not an option for the professors before us. It has
many benefits:
● It allows you to find a community of other professors, get support, form friendships, and
chat with others going through the same things.
● It allows you to broadcast your research easily. Before social media, this was primarily
done by traveling to conferences or other universities

66
 ● On a related note, it allows you to provide your takes on technical events/breakthroughs,
and engage with the community, without publishing papers or attending conferences
● It allows students to identify what kind of an advisor you will be, and therefore can attract
students to your group

Of course, as with any tool, it is possible to use social media badly. You should remember that it
is a public space, with your posts visible both to students and your peers (and potential future
employers). You want to maintain a certain sense of decorum with your official account,
especially since your posts do reflect on your university in a way (even if the views are only your
own).

I primarily used Twitter, and it was extremely beneficial for me. It allowed me to gain visibility for
my group’s research without having to travel a lot. It allowed me to network with other faculty
without having to attend program committee meetings. It allowed me to broadcast to students
what I was looking for, and the culture of my research group. It would be much harder to do
these things without social media.

Overall, a professor doesn’t need to have a social-media presence. It is completely possible to

have a thriving academic career without ever being on social media (as demonstrated by most
senior profs). However, social media offers opportunities that are hard to come by otherwise.
For example, some of the opportunities I got were because some senior professors noticed my
tweets. It would have been hard to replicate this at conferences.

So what should you post on social media? To start with, you can post the highlights. You could
post whenever you have a paper accepted. I usually post something like “Excited to share that
our paper on X got accepted at Y. This is joint work with.. . Read on the main ideas and insights”.
Make sure to tag your students – it is a big deal for them, and helps them get on the radar of
employers. You could post whenever you or someone in your group gets some major award or
grant. I usually learn about grants and awards on Twitter, it spreads fast.

Once you are comfortable doing this, you could branch out a bit more and share more “slice of
life” things on social media. For example, I usually share pictures when I’m out for dinner or a
fun event with my research group. I sometimes share pictures from brainstorming sessions.
These posts help students understand what life is generally like in your research group.

You could post about your reaction or comments to major news events. For example, if a new
technology came out, you could present your take on how it will impact academia or your own
sub-area. For example, a lot of professors talked about the impact of the US administration’s
guidance to make federally funded research open access.

Professors also use social media to discuss research that they read or come across and liked
(or disliked). There have been interesting Twitter conversations about various papers posted on
Arxiv – in fact, many professors use Twitter as one of the main ways in which they discover the
latest research (especially machine learning).
67
Things not to post on social media. This should be pretty obvious, but do not post anything
obscene or vulgar. Always remember that you represent the university, even in your personal
account. One way to remember this is that if you post something terrible and it goes viral, it is
definitely going to be reported as “University of X professor.. “.

On the less obvious side, it is fine to critique the system or a conference's policies, but posts like
“John Doe’s work is terrible” are poor form. There is a fine line here where you can criticize
someone’s work, but not criticize the person. This can be hard to navigate, and if you mess up,
you end up hurting both the person criticized and yourself. I would be careful around this; just
vent in person to some trusted colleagues over coffee or a drink.

Another thing to be careful about is politics. Again, remember that your posts are public. Angry
folks might email your posts to your university administration, making all sorts of complaints
about you. I know some university administrations handle this well, for example the University of
California system. Others might not. If you are posting something political, always ask yourself
if the post were to end up in front of your Dean, what would the consequences be? If you are just
beginning to use social media (and you are also just beginning your academic career), best to
steer clear of this.

Maintaining authenticity on social media. Some students tell me that they feel being on social
media forces them to portray themselves in a certain, inauthentic manner. I agree, but this is
also true of being at work in some sense; your professional self and your personal self are
different. So there are limits on what you can say and post on your professional social-media
account. If you have a separate private account not linked to your professional identity, you are
free to post whatever you like there.

Should you post about your likes and dislikes on social media? Should you let your personality
come through, apart from the strictly curated professional image?

Having a bit of your personal self come through social media makes it easier to find friends in
the academic community who share the same likes. When folks share their personal life, I find
myself relating to them more: ah, that superstar researcher is just somebody like me!

I would say feel free to share anything you could tell peers that you just met at a conference
poster session. Do you love cycling? Great, feel free to share. Do you love baking? This was
really popular during COVID, so feel free to share. Do you love rock climbing and want to talk
about the great gym you found? Go right ahead. Skip anything you would not be comfortable
telling strangers.

68
6.4 Giving talks
One way to spread the word about your research is to give talks at other universities and
companies. This has a number of benefits:
- You get to pitch your idea to folks, and get their live feedback. This can be especially
useful for work that is still in progress.
- You get to meet people who are interested in your work, and this can lead to
collaborations.
- Hearing a talk from the authors is the best way to learn about a work; it is easier than
reading the paper and more memorable. Folks who come to your talks will often later
read your papers and cite your work.
- You get to establish your brand. If your work is thematically focussed on something,
people will remember you later as “person from university X who works on Y”.
- Sometimes the undergrads come to your talk, get inspired, and then want to work with
you for grad school. It is a good way to catch the attention of strong students.

The only problem with this approach is that it is expensive, both in terms of money and your
time. Many professors will do a “tenure tour” just before they go up for tenure, to ensure their
letter writers know their work and have a chance to meet them. I’ve heard this is expensive and
exhausting (but also fun). For my own tenure tour, I gave talks on Zoom due to the pandemic, so
I couldn’t experience this directly myself.

Getting invited. If you are interested in giving a talk at a university, just email someone you know
there and express your interest. Absolutely no one turns down a talk from their peers. The only
catch here is that you must be known to the folks arranging the talk (you must be someone in
their research community). But academic communities are rather small – if you don’t know
someone personally at the university, chances are someone you know does. Request an
introduction to someone they know at the university. Getting a talk invite at a company works
the same way.

Zoom talks. While you might reserve doing the physical talk tour for tenure, I would recommend
doing a similar tour on Zoom for every big paper that you publish. Remember the idea is to have
a few papers that are highly impactful and that you are known for (not a big mass of unknown
work). Zoom talks are easy to set up in that there is no travel involved. The Zoom talk is also a
good bar for deciding whether to work on something: if this work gets published, would you be
excited enough to give a dozen Zoom talks about it? If not, should you be working on it?

69
6.5 Networking at conferences

Networking at conferences is a core skill that you will need to develop as an assistant professor.
Success in academia depends a lot on being invited to things – to program committees, to
journal editorships, review panels, etc. Getting invited is a combination of three factors: the work
that you do, who knows you and your work, and luck (as with everything in life). Networking is
about increasing the set of people who know you and your work, so that the next time they need
an expert in your area, they’ll think of you!

So the first thing to realize is that the best way to attend a conference is not to sit through all the
talks. Really! You might feel like that's the "best value" (given how much you spent to get to the
conference) but it's really not.

The talk marathons that we have in CS conferences are a consequence of our unique
conference-first publishing model, where we publish our work in conferences rather than in
journals. In most other fields, journals are meant for publishing research, and conferences are
meant to meet people. As a result, there is a lot more time planned for interaction.

The way to make the most of a conference is to attend a few sessions, but spend the rest of
the time meeting people.

Go through the schedule beforehand and mark out a few sessions you would like to attend. For
those sessions, pay attention, and go up to the microphone and ask any questions that come to
mind. Speakers love when folks have questions! There is nothing more deflating than practicing
endlessly and giving a good talk only to be met with no questions. So your questions will be
appreciated.

There is this wrong notion that for networking to be useful, you have to talk to
"important/famous people". And because it's scary to talk to famous people you haven't met
before, you get stuck and don't get started. This is closely tied to the notion that you should
have a goal in networking (“to impress Prof. X”).

The goal. Here's what I'd recommend instead. Your goal is to be able to go up to a new person
and have a conversation for five minutes without having anxiety or panic attacks.

That's it. It could be anyone, like a new student at another research group. There doesn't need to
be a "payoff" in the conversation, and you don't need to explicitly ask for anything or say you are
looking for jobs or anything like that.

What should you talk about? Have a short introduction and one-two sentence elevator pitch
(see this from John Wilkes at EuroDW 2018) about your work ready. “Hey I’m Vijay
Chidambaram, an assistant professor from the University of Texas at Austin. I work on storage

70
systems for emerging technologies and applications”. Cap it to a minute or so – give the other
person time to absorb and respond. Ask them about their work. People love to talk about their
work, so this is always a ready topic!

One of the nice things about academic confs is the shared context that provides material for
conversation. Talk about the food. Talk about the keynote. Talk about the poster session!

Obviously, this comes easier to some than to others. It is harder for introverts and shy folks.
Regardless of how naturally gifted you are at this though, it's a learnable skill that gets better
with practice. Once you get better at this skill, you will then be able to approach people you want
to talk to.

Poster sessions and evening events. Make sure you go to the poster session. Look for the stuff
close to your work. Introduce yourself and ask questions! Again, authors love when folks come
and ask questions at the poster session! If there are other evening events, attend those if you
can. Sometimes there are fun Ask Me Anything or similar sessions later in the day.

Taking a break. Feel free to take a break and recharge as much as you need. Weaving in hallway
sessions and talk sessions gives you a change and a break from each type of activity. I'm
usually exhausted at the end of each conference day. It can be exhausting having to be "on" all
the time. This is normal, and okay. Get a good night's sleep!

To summarize, attend a few sessions, and spend the rest of the time in the hallway. Practice
introducing yourself and talking to new people.

"Success" in networking just means meeting people, learning about their work, and spreading
the word about yours. You don't need to have tangible outcomes like a new collaboration or
interviews or something like that for it to be successful. Focus more on the process (being able
to introduce yourself, holding a conversation) than the outcomes. You will rapidly get better at
this, and as the set of people you know grows, it will become easier at the next conference. With
time, you won’t have to consciously do this – conferences will just become a place where you
meet friends and talk about the cool work your group is doing.

Other Resources

Check out this excellent advice on attending conferences from Michael Ernst and David Notkin.

Summary
We have talked about how to do outreach so that the world learns about your
awesome work. Outreach is a crucial component of being successful, especially
when you are an assistant professor without an established brand. Don’t be shy
about using every tool you have to do it! Next, we will talk about service for your
academic community.

71
Chapter 7: Service
One part of the professor job that doesn’t get a lot of attention is service. However,
understanding and managing this component is important, both for your own tenure case and
for the wider community. Let us start by talking about what service is, and why it is important.

7.1 Defining and managing service

So what do I mean when I say service? I would loosely define service as optional tasks that your
community requests from you. You are free to decline these tasks without a direct impact on
your career (there might be indirect impacts). These tasks are important to keep the community
running.

For example, your peer researchers may ask you to serve on the program committee for a
conference, or review a paper for a journal. You will get a lot of these requests, and you are not
expected to say yes to all of them. Your department will expect you to do some service each
year, and there is an expectation that you get invited to serve on the PCs of the top conferences
in your research area. But beyond that, you don’t get any benefits by doing a lot of service.

Why should you do service? To put it in a nutshell, doing service is part of being a good citizen,
both inside your department, and inside your wider academic community. This is because of the
way academia operates, fully dependent on freely volunteered labor.

For example, imagine that you want to run a conference. You need someone to organize the
conference, and someone to review submissions and select papers for the program. One way
to do this is to recruit reviewers, pay them for their time, and pay other folks to handle the
logistics of running the conference. Speakers would get paid. All the money would come from a
combination of sponsors and the fees that attendees provide. This is how industrial
conferences operate.

Academia uses a different model. People volunteer their time to act as reviewers or conference
organizers. This allows the conference organizers to reduce the cost that needs to be paid by
attendees (even so, the cost is usually several hundreds of US dollars). USENIX uses a slightly
different model where the conference organizers are paid professionals, but the technical
reviewing is still done by volunteers. This results in a slightly higher registration fee for
attendees (compared to when everything is done by volunteers).

Thus, almost everything in academia runs on volunteer labor. All the events you attend, the
conferences and journals you publish in, the panels you enjoy, the awards that are given out, all
depend on people volunteering their time. Since you get a net benefit from the time volunteered
by others, it is only fair that you do your share of volunteering as well.

72
There are some other fringe benefits to volunteering. One is that you get your name out. When
you are an assistant professor and you are struggling to establish yourself, it can be useful to
have folks know you and recognize your name. The second benefit is that you get to meet folks
in the course of doing service and form your network; this is particularly useful when you are
just starting out.

Why should you not do service? While it is important to do service, I’ve seen folks go overboard
on this (and did too much myself in my first year). Doing service feels good: you are doing
something for your community, and it feels like you are getting your name out. And it is hard to
say no when folks you respect request you to join a committee.

But you should always, always remember: you will not get tenure for service. No matter how
good the service is. The first consideration at R1 universities is always research. If your research
is not good, having done amazing service will not get you tenure.

Apart from not helping you with tenure, taking on too many service commitments will make you
bad at service. If you say yes to too many concurrent PCs, you will have a tough time doing a
good job at reviewing papers for all of them. Nobody wins in this situation: the conferences and
authors get crappy reviews, and you get a reputation for writing bad reviews.

Managing service commitments. As an assistant professor, you should do enough service to be

a good citizen, but not so much that it affects your other job responsibilities. You should
conserve your limited energy for research and teaching. Be thoughtful and intentional about the
service requests you accept – always think about how much time you are committing down the
line (even if it is several months or a year away). The way to think about it is that you can do two
to four big service requests each year. A big service request is something like an organizational
role at a conference, or a program committee assignment where you will be reviewing 10+
papers. You should really be doing only a few of these each year. Smaller service requests, such
as being an external reviewer for a conference, you can do more of; but be careful that they don’t
pile up.

Saying no. How do you say no politely, especially when it is someone you respect and look up
to? You should remember that folks who run conferences expect a lot of people to say no. It
happens all the time, and if you decline, they will not take it personally. You don’t need to worry,
“If I say no this year, maybe they will not invite me ever again”. I can relate because I had this
exact worry at the beginning. I’m happy to report that I have declined to review at conferences,
and still been invited the following year. As long as you do good research and write good
reviews, you will be invited back. So do not let this consideration make you say yes when you
don’t have the time to do a good job.

I usually say something like: “Thank you for inviting me. Unfortunately, I am overloaded on
service commitments this year, and I don’t think I will have enough time to do a good job.
Therefore I will have to decline”. As you get a bit more senior, you could recommend a junior
faculty member they could invite in your place.
73
7.2 Serving on program committees and editorial boards
When you get started as an assistant professor, serving on program committees will form the
bulk of your service. At first, it can be super flattering to be invited to a program committee: I
definitely remember my first such invitation. You now have a voice at the table in determining
the program! You can review papers alongside researchers you looked up to. As a result, you
pretty much say yes to all the invitations you initially get.

This is not a great strategy, though: you end up over-committing yourself in your initial years. I
would recommend saying yes only in the following cases:
1) The conference is your home venue, where you have been publishing in grad school, and
where you hope to publish in the future.
2) It is the top (or one of the top) conferences in your area.
3) It is a community that you want to join, or an area that you want to publish in.
4) It is a workshop specializing in your area.

For example, I publish regularly at systems conferences such as ATC/FAST, so I tend to review
papers for them. SOSP and OSDI are the top conferences in my area, so I almost always accept
PC invitations for them. HotOS and HotStorage are the relevant workshops for my area, so I
review papers for them as well.

Don’t accept invitations to be on the PC for conferences that you have never published in, and
are unlikely to do so in the future. Don’t accept invitations just because you know the PC chair if
you are already committed to other PCs.

Overall, a good number to aim for is about 1-2 big PCs each year (10+ papers per PC) and 1-2
smaller ones (5+ papers per PC). Between research and teaching, it will be hard to do a good job
with more PCs.

University expectations regarding PCs. Generally, your department or university likes to see that
you are being invited to serve on the PCs of the top conferences in your area. If you become an
assistant professor and receive no PC invitations, that can be a bit of a red flag. However,
beyond expecting to see some service (a few PCs each year), the department/university doesn’t
really care how much service you do – they assume you will manage your service workload
efficiently.

How much time does reviewing a paper take? This really varies by area. I would say anywhere
between one to four hours per paper is common. Do not spend more than half a day on a single
paper. You should read actively, trying to get the main ideas and insights, which will enable you
to write a good review. You should aim to read the paper only once (this gets easier with
practice), taking notes as you go. Having questions you want to answer as you read will help
make reviewing easier.

74
If you find yourself spending more than 1 day a week on average (or 8 hours per week) for
reviewing papers, you have accepted too many PC invitations. It can be hard to properly track
this though, since you have a lot of activity near the reviewing deadlines, and not much on other
weeks.

I’ve mostly talked about program committees since conferences are considered the primary
publication venue in computer science, but much of the advice applies to journals as well. Only
review for journals where you publish regularly, and track and manage how many submissions
you accept to review.

7.3 Serving on dissertation committees

Another form of service is participating on dissertation committees, where you read the
dissertation and attend the thesis proposal/defense. This can take up a lot more time, since you
have to read an entire dissertation. It helps if you already know the students work, since that
will form a significant portion of the dissertation. Still, serving on a dissertation committee can
be an entire day’s worth of work.

Requests for serving on dissertation committees can come from within the department or
externally. The internal ones are harder to say no, since these are the colleagues you work with
on a daily basis. Nevertheless, track how much time you are spending on this. Don’t say yes to
more than one or two dissertation committees per semester/quarter.

From the university viewpoint, I don’t think there are concrete expectations regarding serving on
dissertation committees. Not being on dissertation committees early on in your career will not
be seen as a red flag (to the same extent that not being on PCs would be).

7.4 Serving as a session chair

Another form of service is serving as the Session Chair at a conference. The session chair will
introduce the session (typically 1 to 1.5 hours) at the front, introduce the speakers, and
moderate the questions. The session chair also kicks off the QnA if there are no questions from
the audience at first.

Being a session chair is a fantastic service opportunity, and you should definitely sign up
whenever you get the chance. It gets you face-time in front of the community, you get your name
out, and the work required for being a good session chair is modest. Grab session chair
opportunities with both hands when you are starting out!

7.5 Organizational service roles

As you become a bit more experienced, you will be offered organizational roles such as being
the Publicity Chair, the Web Chair, the Poster Chair, or the Program Committee chair. Of these,
being the chair of the program committee is most prestigious, since you are responsible for
75
delivering a high quality program in coordination with your chosen program committee. So if you
are invited to be the PC chair of a conference or workshop you normally publish in, you should
accept the invitation. Your university or department will care about whether you have been
invited to be the PC chair (once you become a bit more senior).

However, the other roles involve a lot of work and little recognition. Ideally, these would be done
by professionals (as in the case of USENIX). However, for non-USENIX conferences, someone
has to step up and do them. Be aware of the work-reward ratio if you choose to accept an
assignment like this though.

During my time as an assistant professor, I co-chaired HotStorage 2020 and SyStor 2021. I was
also one of the General Chairs for SOSP 21. I enjoyed doing all three roles, but it was definitely a
lot of work! I would not recommend doing any of these roles (especially being General Chair)
early on as an assistant professor; it is a better fit when you are close to going up for tenure.

7.6 Departmental and university service

Apart from serving on program committees, editorial boards, and organizing conferences and
workshops, you are also expected to do “local” service for your department and university. This
can take multiple forms.

At the department level, this means you will serve on one or more committees. Assistant
professors are usually assigned to either the PhD admissions committee or the faculty
recruiting committee. Both committees are extremely important and directly influence the future
of the department. Other committees include Faculty Evaluation Committees, Budget
Committees, and Promotion Committees that you can participate on once you become more
senior. Expect to spend at least a couple of weeks (put together, it will be spread over the
semester) on departmental committee work each year.

At the university level, there will be committees deciding overall university policy. Typically, your
department or college will need to be represented on these committees. As an assistant
professor, it will be rare for you to be nominated by your department for these committees;
typically, more senior professors participate in these committees.

Summary
The most important thing to remember about service is that it is like chores around the house;
you need to regularly clean your house, water your plants, etc. You do this because you want a
nice place to live in; similarly, service provides you a nice community to do research in. However,
carefully manage the amount of service you do. You will be bombarded with requests, and it is
all too easy to say yes to too many requests. Don’t accept so many service requests that it
impacts your other responsibilities as a professor.

76
Chapter 8: Self-Care
In this last chapter, let us talk about self-care while doing the assistant professor job. This is a
hard job for many reasons: the hours are long, there is a lot of ambiguity about what to do and
how to do it, and you have to play many roles to do the job right. If you are not careful, the job
can take over your life, eating into time with family and friends, leading to stress, depression,
and burn out. We will talk about how to manage these aspects of the job.

8.1 Don’t give up health for tenure

I think a temptation for many folks starting as professors is to focus on the job at the exclusion
of everything else: “let me figure out the job, I’ll take care of my health afterwards”.

I would not recommend this approach. The professor job is more of a marathon than a sprint:
there is always another paper to write, another grant proposal to put together, another
collaboration to participate in. If you put off taking care of your health until you have some
breathing space, it will never happen.

What I recommend instead is figuring out a stable, sustainable work rhythm that includes time
for health, friends, and family. In Chapter 1.7, I talked about what my routine looked like: work
begins at 9:30 - 10 am in the morning, goes until 5 pm, then time for gym, friends and family,
then work again for an hour or two after 9:30 pm. I normally do not work on the weekends. Of
course, nearer deadlines, things become more hectic, but that is usually limited to one week or
so near the deadline (and only 3-4 such deadlines through the year).

Your goal is to stick to your routine regardless of how research or teaching is going. You want to
get more efficient at using the work hours in your routine, without requiring to borrow time from
friends or family or your health. This is something I saw even in grad school: the most
productive folks do not depend on temporary “bursts” to get their work done. There is usually
such a push near the deadline, but the majority of the work gets done at a steady pace.

The benefit of the routine approach is that it is low stress. Because you have predictable work
hours, you can plan ahead and allocate time for research deadlines, teaching, and service. Work
in time to get exercise, and to have enough time to eat healthy food (it is typically easier and
quicker, but not healthy, to eat fast food).

Over the long run, your research career is going to depend much more on the body of work that
you produce, rather than on any single result/paper. Thus, rather than producing one paper and
burning out, it is more useful to figure out how to produce a series of work sustainably.

77
8.2 Making time for family and friends
There can be a temptation to work through evenings and weekends, especially at the beginning,
to ensure you are in a good position for tenure. I would caution against this. I think working
evenings and weekends during the occasional deadline is fine, but it should not be the common
case.

The reason not to do this is burnout. Being a professor is a job that requires a lot of motivation.
You don’t have a boss – if you decide to slack off, there is no one to tell you not to do it (until it
is too late). Being a professor means that you have to be internally driven to do the next project,
to write the next paper, etc. Burnout makes you lose your motivation, and shortly thereafter your
productivity falls off. Make sure you spend enough time with friends and family to recharge
yourself every week.

This means zealously guarding your off-work hours from meetings (they tend to creep past 5
pm) and your weekends from any meetings or work. Be firm in declining meetings: “Sorry, I am
not available at that time”. You don’t need to provide excuses or explanations.

Similarly, decline any work folks try to schedule on the weekend. I vividly remember one meeting
where a senior collaborator wanted us to do some work over the weekend; a junior collaborator
on the team flatly declined, “Sorry, I will be spending that time with my kid”. That person was my
hero, since the idea of declining hadn’t even occurred to me. I have declined a lot of meetings
and work since then!

A lot of folks will want your time, and they will always make it seem like it is urgent. You should
use your own judgement in deciding if it is really urgent or not. What really happens if the work
is done on the next weekday instead? Oftentimes nothing.

The prof job can seem all encompassing, but we have to remember it is, in the end, a job: a
means to support yourself and your family. Make sure you don’t lose out on other things as you
pursue tenure.

8.3 Dealing with peer pressure and comparisons

One of the best things about academic life is that you are surrounded by amazing, smart,
motivated colleagues; this is also one of the worst things about academic life, due to peer
pressure and comparisons.

One of the things you might need to accept as you start your job: there will always be colleagues
who are more successful than you. If you have a need to “be the best” (which can be a good
thing, and part of what brought you to the job in the first place), you might struggle a bit. Every
day, someone is winning an award, someone is receiving a grant, someone else is getting their
paper featured in the New York Times.

78
Such comparisons can devour you if you let them. I typically try to focus on my own goals, my
own research group, and what I can do to achieve my goals. This brings the focus back on
things I can control, rather than making unhelpful comparisons. I try to remember that as long
as I am doing research that I like and find interesting with my students, and I am able to publish
and fund my students, that is all that matters at the end of the day. When I retire, it is unlikely
that I will remember my awards; I will remember all the relationships that I built with my
students and colleagues, and how much fun it was on a day-to-day basis.

8.4 Creating a support network

I would strongly recommend having a support group of other assistant professors, both at your
own department/university and elsewhere. For me, this naturally turned out to be the folks who
started as professors at the same time I did (my “cohort”). Among profs senior to me, Isil Dilig
was an amazing mentor, always willing to listen and provide feedback.

Having a support group like this, with whom you can talk about the job and rant about all its
problems and things that don’t go well, is crucial to maintaining sanity as a professor. For
creating a support group outside your university, use social media. Twitter was extremely useful
in this regard, as I got to know professors at various universities who I had never met in real life,
but nevertheless shared a lot with.

8.5 Dealing with rejection

Academic life is filled with rejection. We constantly submit papers and grant proposals, only to
have them shot down. Most venues and funding opportunities are competitive, with a 10-25%
success rate. This means that most of your efforts, especially early efforts in a new venue or
funding opportunity, will fail.

Having a paper get rejected after a lot of work can be tough. Similarly, working hard on a
proposal draft, only to see it getting rejected months later, can be devastating. It is important to
handle these rejections well.

One of the things that helps me deal with rejection is that every researcher has to go through it.
I’ve collaborated with amazing researchers, and they all get routinely rejected. It is simply a fact
of academic life (at least, as we have constructed the system for now).

It is important to separate the rejection of the submission, from rejection of the work, and
rejection of you as a researcher. Most rejections are just about the particular state of the draft
submission; it has little to do with the core idea itself (and nothing to do with the authors). Many
pathbreaking results that we now take for granted were rejected initially – even Nobel-prize
winning work has been initially rejected. So a rejection just means that the current draft is not
good enough; perhaps the story has to be told differently, perhaps more experimental evidence
is needed. This is something you get better with practice over time. “A draft I wrote wasn’t
persuasive to reviewers” is much easier to take than “they rejected my work” or “they rejected
79
me”. Be careful in how you think about rejections, even to yourself (perhaps especially to
yourself). Once you start thinking about drafts being rejected, it becomes easier to take and to
move forward.

Develop a rejection routine. I recommend developing a routine to handle rejection. For example,
when I get the decision from a conference or grant agency, and it is “Unfortunately,..”, I just stop
reading the email. Rejections almost never require you to take immediate action. I then go get
an ice cream or coffee with my students (who will also be similarly bummed out) or a drink with
my colleagues (if we are sharing rejections). It is super effective to get together with other
colleagues who have also been rejected and just talk. I watch some of my favorite go-to happy
movies to feel better.

After a couple of days, come back and read the reviews. With a clearer head and hopefully less
emotional attachment, you can figure out if there is anything constructive in the reviews that you
can use. Sometimes there is, and sometimes there is not. Note that in many cases the reviewers
who rejected your paper are not the reviewers who will see the next version; you want to avoid
overfitting to those particular reviewers. You should get together with your collaborators and
figure out what is next for the paper or the grant proposal.

8.6 Having kids while on the tenure track

I started the professor job in 2016 (when I was 28). My wife and I had our first kid in 2018, and
our second kid in 2022 (just before tenure). I know successful colleagues who had multiple kids
when on the tenure track as well.

I think doing well academically after having a kid depends on two things: whether you have
some family support to take care of the kids, and whether you are able to find good daycare.
You would think finding daycare is a solved problem in this day and age, but it is sadly not. I’m
lucky that daycare costs in Austin are reasonable (~$1000 per kid per month); in places like
Boston and Seattle, it is around $2500-$3000 per kid for each month. It is extremely hard to
support a family on one academic salary.

Having a kid introduces a lot of additional constraints on your time. When my first kid came
along, it altered my post 5pm working hours a bit, since I had to work while he slept. When he
got older, I would have a hard stop at 5 pm, since I would need to go to pick him up at daycare.
The days of parents are basically structured around their children, their daycare, and when they
go to sleep.

Having kids made me more efficient. I learnt to do more within my work hours, and to ensure
meetings got done on time. It also made work-life balance a bit easier, since you can say no to
any meetings in non-work-times without feeling guilty.

80
8.7 Speaking truth to power
One of the unique benefits of being in academia is that, in some places in the US, you can
directly and publicly criticize your employer without fear of being fired. Of course, this protection
is much stronger once you have tenure, but in general, professors have much more leeway in
what they can say publicly about their employers. For example, if you were disappointed with
the way your university handled the pandemic, saying so publicly should not result in loss of
employment.

As such, professors have a strong voice in their communities. This voice can be used to call out
problems and bring attention to issues. You can use it to amplify others' voices. If you use it
properly, it can be a significant aid to those without power in your community.
Of course, the safe course of action is not to say anything publicly. Taking a side on some
issues will inevitably earn you enemies. For example, if you come out strongly in favor of
diversity, equity, and inclusion initiatives, there will be some who will take offense to that. But
you will also gain allies that you would not otherwise. For example, Arun Kumar at UCSD was
actually “praised for his talent for being outspoken”.

Summary
Being a professor can be stressful and overwhelming. Make sure you make time for your health,
for family, and friends. Create a support group of professors who you can chat candidly with.
Have a routine for dealing with rejection. In the end, remember that it is just a job, meant to
support you and your family!

81
Conclusion
Academia can often be opaque to outsiders. I hope this book shows you what being a professor
at an R1 university in the USA is like. Being a professor is not for everyone: you give up
significant income, and perpetually make a fraction of what you could get in industry. However,
being a professor also brings you joys and freedom that is hard to obtain in most other jobs. If
you are considering becoming a professor, I hope this book allows you to make a more informed
decision. If you are starting soon as a professor, congratulations, and I hope this book helps you
a bit on your academic journey!

I would love to hear your thoughts on the book. Please email vijayc@utexas.edu with any
comments or questions.

82