You are on page 1of 19

DATA DIGEST

BITS & BYTES


archushukla123@gmail.com
BK8EHPNQZI

APRIL 2023 EDITION

This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
WHAT’S

INSIDE?
Leadership Speaks 03

Great Learning Journey 04

Discover 06
archushukla123@gmail.com
BK8EHPNQZI
That’s A Good Question! 07

What’s New? 10

Industry Trends 12

Data Science at Work 14

AI at Work 15

Mentor Speaks 16

Crossword Puzzle 18

This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

Q2. What advice would you give someone


LEADERSHIP who is about to handle a leadership position

SPEAKS
for the first time?

• Build self-awareness

• Never deviate from your core values

• Think about other people’s motivations


comprehend them and then factor in on
why people respond the way they do /
might do

• Prioritise actions based on the impact

• Don’t be afraid of conflict but be respectful


and decent when you have a difference of
opinion

Q3. How do you respond to criticism?

ANIRUDH LUTHRA • Filter the message to only include what is


Associate Director, actually being spoken - this suppresses the
archushukla123@gmail.com emotional response
Operations, Great
BK8EHPNQZI Learning
• Reflect and look back to see if there is a
Q1. What are your core values? How do you trend where something similar has come
ensure that the organisation and its activities up before
are aligned with them?
• Make a mental note in the self-awareness
1. Do the right thing for the right reasons bucket
(for our customers, our colleagues and our
business) • Be fair to yourself, you are only human:)

2. Ensure people get credit for their ideas Q4. Talk about a leader that inspires
and efforts you and why?

• I repeatedly talk about these when I I am actually inspired by two sportsmen, MS.
interact with my team and colleagues Dhoni (Cricket) and Ben Hogan (Golf). One of
whom is a leader and the other a pioneer of
• I constantly remind people around me of
his sport.
why we do what we do and the impact
our actions can create in the professional Their journeys to leadership are of persistence,
and personal lives of others. Be it our resilience and dogged determination. The
customers or individuals in our immediate impact that they have created will last for
vicinity years (Ben Hogan fundamentally changed
how people approach the game. The
• We must acknowledge when people do the
leadership mindset and changes to cricket
right thing
that MS has brought will continue for years).

3 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

I went through the entire medical system last


GREAT LEARNING year (2021–2022). Soon, I discovered that

JOURNEY millions of lives could have been saved if AI


was used to improve the decision-making
process for diagnosis and treatment. That's
when I learned about Great Learning from a
friend.

The program structure is excellent. It teaches


topics that you will need when you enter the
market. The mentor learning sessions were full
of high-quality presentations.

I was only able to take a deep dive into the


area of Data Analytics because of Great
Learning. I salute the high standard of
delivery of the Mentors I have worked with.
The Program Managers were always willing
JOY CHANDRA to assist. They always worked with me, even
when I had health problems or office work
PGP-DSBA PROGRAM ALUMNUS
pressure.
archushukla123@gmail.com
BK8EHPNQZI
The PGP-DSBA Program from the Great
Everyone has a unique tale to tell. Mine is very
Learning has sparked a research analyst in me
distinct. I wanted to master Data Science and
who views the world from a different angle
contribute to the huge field of implementing
and enjoys applying his knowledge when the
Data Science, Machine Learning, and Artificial
need arises. I continue to seek chances to
Intelligence in cancer detection and therapy
obtain a Ph.D. in analytics.
because I had colorectal cancer. Cancer is
an aggressive illness that costs us money, I believe, while having a fundamental
but it can be treated if caught early. I had to understanding of logic in coding is necessary,
learn how to use AI and Machine Learning the ability to comprehend domain-specific
in business applications that could help our data is ultimately what will determine if
societies combat cancer. someone can master analytics.
The PGP-DSBA Program at Great Learning
is a perfect synthesis of Data Science and
business.

I currently work for Ericsson as a telecom


professional. I am experienced in automating
cloud network deployment for 5G telecom.
I've always wanted to apply my virtualization
expertise to analytics, particularly in the field
of healthcare.

4 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

statistics and probability were important.


GREAT LEARNING These are some concepts that we learn in

JOURNEY school, so suffice to say that I was out of


touch.

Also, pursuing an intensive program like this


with a full-time job was a big challenge, but
with the help of my Program Manager, peers
and mentors, it all worked out just fine.

ML and Data Science are one of the hottest


topics in the world right now, hence they have
better opportunities. Also, I grew old from
my previous profile, felt stagnant, and after
a point was doing it just because it paid the
bills, not because I liked it. I knew I had to
keep up with the pace of the world. That’s why
I was desperately looking for a transition into
SUNAAL DUA this field.
PGP-AIML Online Course Alumnus
The program operates on its own since it
archushukla123@gmail.com
BK8EHPNQZI is so nicely designed. After I finished my
foundational coursework, I was able to
I have completed three years in IT and have
quickly grasp complex ideas in ML and Data
recently transitioned into Analytics. I am
Science because I had a solid foundation in
currently working as a Data Engineer and
mathematics and algorithms.
Analyst with a progressive organisation
shouldering Data Transformation and My mentor was also amazing. He went out
Automation pipelines. Before PGP-AIML of his way to help me and solved my queries
Online Course, I was working as a Backend outside the curriculum as well. He also helped
Web Engineer for approximately 2.5 years. My me whenever I got stuck with my project.
technology stack before pursuing the program Even after the program, we are still connected,
was: and I still ask him whenever I get stuck
PHP-Laravel, Python-Django, MySQL, somewhere. Great Learning is a great blend of
JavaScript and Ajax. mentorship, theory, and practice. There were
times when I required extra studying materials
I knew for a fact that prior experience in data
and my program manager actively helped in
modeling and management is a must in the
providing them.
Machine Learning (ML) or Data Science space,
which I was missing at the time. I never got Finally, I was able to make the desired
the chance to deal with big data as a backend transition with the help of the career
web engineer. So learning SQL and managing assistance provided.
big data was mandatory before/during the
program. At the same time, concepts of For people who are starting this journey,
I would suggest that they learn things with a
mathematical perspective, not just intuition.

5 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

DISCOVER
AI has been a game-changer across diverse
industries and the sports industry benefited
from the same. With applications across
in-game activity, post-match analysis, and
ChatGPT for Data Scientists even fan experience, AI has made the game
What if you could instantly understand more engaging and enjoyable.
archushukla123@gmail.com
vast amounts of data with just a few simple
BK8EHPNQZI
prompts? What if you could extract insights The future of sports undoubtedly lies with AI
and generate predictions with uncanny technology. It does not alter any outcomes but
accuracy? Enter ChatGPT – a language model assists in making effective decisions. All these
created by OpenAI that has the potential examples and technological aspects are just
to revolutionise the way we approach Data the beginning, and there is more that is yet to
Science. By training on massive amounts of be offered by AI which now heavily banks on
data, ChatGPT can generate natural language the rise of Machine Learning.
responses to prompts, making it an incredibly
versatile tool for various Data Science tasks. Technology is indeed an exciting domain.
As the day progresses, new technologies
come into the foray and if you want to stay
KNOW MORE
ahead in this interesting domain, you must
keep upskilling with the latest trends. So,
if upskilling is on your mind, check out the
various upskilling programs available at
How AI Transformed the Great Learning.
Sports Industry
KNOW MORE
The use of technology in the sports industry
has risen significantly. From ticketing systems
to player statistics, Artificial intelligence
(AI) has significantly increased the rate of
audience engagement and game strategies.

6 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

THAT’S A GOOD Stemming


Stemming is the practice of eliminating the

QUESTION! final few letters of a word to produce a shorter


variant, even if that form is meaningless.
Words like "historical" will become "histori"
In this edition, our question will be when stemmed. Even if these forms are
archushukla123@gmail.com
BK8EHPNQZI illogical from a linguistic standpoint, we
Difference Between Stemming can nevertheless utilise them for sentiment
analysis, spam detection, restaurant
and Lemmatization in Text evaluations, and other applications. It is crucial
Analytics? to determine if the root word in such problem
statements is positive or negative. This base
Gaurav Das, a Great Learning mentor says word can be derived via stemming.
Lemmatization and stemming are significant
text normalization techniques used in NLP. Lemmatization
Lemmatization is the act of reducing words Lemmatization serves the same objective but
to their base or dictionary form, whereas avoids the stemming issue. For some words
stemming is the process of returning such as "histori," stemming may not produce
inflected or derived words to their root form. a meaningful representation. However,
Languages differ in their degree of inflection, lemmatization will carefully extract the right
and some contain more inflectional forms than root word in this instance, which is "history".
others. For instance, compared to English, the As a result, it will take longer than stemming
German language has a significant degree of because it discovers the right form of the
inflection and necessitates more sophisticated root word. Stemming only shortens the input
stemming and lemmatization procedures. word to produce a base word, which reduces
Porter Stemmer and WordNet Lemmatizer processing time.
are two functions from the nltk.stem library
in Python that can be used repeatedly for
stemming and lemmatization.

7 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

Thenkavi Elumalai a Great Learning mentor


While lemmatization is used in more
says
sophisticated and specialised fields like
chatbots and human answering, stemming is Natural language processing (NLP)
used in sentiment analysis. procedures like stemming and lemmatization
are used to normalise text and get words
Stemming and lemmatization each have and documents ready for Machine Learning
advantages and drawbacks. processing.

The benefit of stemming is that it can greatly For instance, in NLP, it is important to
minimise the quantity of unique words that understand that the verb tenses "wish" and
an algorithm needs to process, which can "wished" refer to the same word. The next
enhance the efficiency of the algorithm. step is to stem or lemmatize both words in
Consequently, it becomes simpler to evaluate, order to get them down to a single root. In
contrast, and comprehend texts. This is useful this approach, the two words are treated
for less complex jobs where the objective is identically; otherwise, the model would treat
to ascertain the sentiment of a document, "wish" and "wished" differently than it does
such as sentiment analysis or document
classification. However, the issue of
over-stemming still exists, where two different
words (for instance universe and university)
can produce the same output (univers).
archushukla123@gmail.com
Under-stemming is the term for situations in
BK8EHPNQZI
which the stemming of two different word
forms yields totally different stem outputs.

The obvious benefit of lemmatization is the


extraction of actual dictionary words from
the input word. However, this procedure may
take some time. It results from morphological
analysis and determining a word's meaning
using a dictionary.

It is safe to say that lemmatization and


stemming are both crucial NLP techniques but
which one to use depends on the task at hand
and the language being processed.

8 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

"wish" and "car". such as the stemming process not working for
words that are not in the English language.
Stemming
In order to establish a so-called root word, Lemmatization
stemming is the act of eliminating suffixes Lemmatization, as opposed to stemming,
from words. The words are reduced to their applies a morphological analysis to words by
base form by a rudimentary set of criteria. taking into account a language's entire lexicon
in addition to word reduction. Lemma for
For instance, the term "run" can be used as "was" is "be," while the lemma for "mice" is
a synonym for all three words: "run", "runs", "mouse."
and "running". By doing this, an NLP model
can discover that the three words share some Most people believe lemmatization to be
characteristics and are used in comparable far more informative than simple stemming.
contexts. The past, present, and future tenses Lemmatization does not classify sentences;
of the nouns are broken down into their root instead, it analyses the context of a word to
forms. Stemming hence avoids the emergence determine its part of speech.
of several columns and reduces it to a single
column. Stemming does have disadvantages.
Stemming might always yield “saw” when
In many applications such as clustering or given the token “saw”, whereas lemmatization
text categorization, stemming enables us
archushukla123@gmail.com might always return either “see” or “saw”
BK8EHPNQZI
to standardise words to their base stem depending on whether the token was used as
regardless of their inflections. Regardless of a verb or a noun.
word form, search engines heavily rely on
these strategies to deliver better results. For instance, the phrase "I am meeting a
person" is a verb where the lemmatization
One of the most well-known stemming process changes it to the root form of "meet."
techniques is Porter's Stemmer Algorithm, I have a meeting tomorrow. Since meeting
which was first suggested in 1980. It is is a noun in this sentence, the lemmatization
predicated on the notion that the suffixes in process keeps the word meeting in its original
English are composites of smaller and simpler form.
suffixes. It is renowned for its quick and easy
processes, but it also has certain drawbacks • Stemming is a quicker procedure than
lemmatization since it removes the word
without regard to context, whereas
lemmatization depends on context.

• Lemmatization is a canonical dictionary-


based approach, whereas stemming is a
rule-based one.

• Lemmatization is favored for context


analysis however stemming is advised
when the context is not significant since it
is more accurate than stemming.

9 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

WHAT’S What is General AI?

NEW
Strong AI, commonly referred to as Artificial
General Intelligence (AGI), is still a futuristic
idea because it includes a computer that can
What are the different types of AI? understand and carry out a wide range of
archushukla123@gmail.com tasks based on its acquired knowledge. As AGI
Three commonly recognised subcategories of
BK8EHPNQZI computers would be able to reason and think
AI—narrow AI, general AI, and super AI—can like a human, this sort of intelligence is closer
be distinguished.
to that of the average person.

What is Narrow AI?

Voice assistants like Siri, Alexa, and Google


Assistant depend on Artificial Narrow
Intelligence (ANI). This category contains
intelligent systems that work without being
specifically programmed to do so, and have
been created or trained to complete particular
jobs or address certain issues.

Since ANI lacks general intelligence, it is


sometimes referred to as weak AI, but some
examples of the power of narrow AI include
the aforementioned voice assistants, image-
recognition systems, tools that identify
inappropriate content online and technologies
that respond to straightforward customer
service inquiries.

10 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

ChatGPT, Deep Learning, Visuals: AI you employ complies not only with legal
Top Do’s and Don’ts of AI in 2023 requirements for copyright compliance but
also with your organisation's own policies and
AI is still being used widely in our daily lives, standards.
whether it is for entertainment, shopping,
healthcare, or other purposes. Despite the
fact that AI can be quite helpful, there are
occasions when it is preferable to avoid using
it. It has also solidified its position in marketing
and publishing.

Dos
#1: Do use AI in video
One of the most reliable applications of AI in
content development is video. Because they
are merely adding pictures to something the
author has already written.

#2: Do use AI for simple automation


archushukla123@gmail.com
BK8EHPNQZI
Anything that gives people more time to be
creative is one of the best uses for AI. Many
laborious or complicated processes in the field
of content creation could be made easier by
machines.

Don’ts
#1: Don’t use AI to write original content
The majority of AI-produced long-form
material lacks the coherence and flow that
only humans can offer.

#2: Don’t forget to double-check AI’s work


Just because AI is a machine doesn't mean it
can't make mistakes. Check the results that
AI produces in the same way that you would
before publishing your own writing.

When we hand that process over to a


machine, it can be difficult to remember
precisely how many components humans
are continually weighing. Make sure that any

11 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

INDUSTRY 1. Developing the vocabulary, the many

TRENDS
categories and the production guidelines
by giving books to GPT-3. The production
rule is produced once the model
Rakesh Lakkala, a Data Science determines the category that each word
archushukla123@gmail.com belongs to.
industry expert, has shared
BK8EHPNQZI
some insights on ChatGPT 2. Generating sentences to feed the model,
resulting in the creation of a vocabulary
The majority of you have probably heard of
and production rules for each category.
"ChatGPT," but what is GPT? Generative Pre-
The model should identify the category of
Trained Transformer is referred to as GPT.
each word in each sentence, and then a
GPT-3, the third-generation GPT, is used by
ChatGPT. GPT-3 is a neural network model rule should be developed.
that was trained utilising data from the Anything that has a language structure can be
internet. OpenAI, a research company made using GPT-3. It can take notes, translate
co-founded by Elon Musk, developed GPT-3. languages, write essays, summarise lengthy
GPT-3 produces text utilising pre-trained materials, provide answers to queries, develop
algorithms that have already received all of computer codes and even discover problems
the information needed to complete their in existing codes. In an online demonstration,
work. Around 570 GB of textual data from it is shown how to use a plugin for the
internet searches and other texts chosen by software program Figma which is frequently
OpenAI, such as Wikipedia were fed into the used for app
system. creation. The aim was to create an app that
Without supervision, the GPT-3 model can resembles the Instagram application in terms
produce texts with up to 50,000 characters. of appearance and functionality.

There are two phases in the GPT-3 neural


network.

12 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

The use of GPT-3 comes with a variety of


difficulties. Computing power, price, system
integration, scalability, and customisation
are a few of these. GPT-3 is a pre-trained,
non-constant learning model. GPT-3 only
accepts small input sizes. We are unable to
supply a lot of text for the result. GPT-3 has
a token cap of around 2,048. The model for
GPT-3 takes a while to produce results. Since
GPT-3 uses neural networks, it is unable to
explain or interpret why particular inputs
result in particular outputs.mean it can't make
mistakes. Check the results that AI produces
in the same way that you would before
publishing your own writing.

How ChatGPT Operates?


As we saw in the last section, the GPT-3
Model can be used to carry out a variety of
archushukla123@gmail.com
activities once it has been tracked. Finally,
BK8EHPNQZI
supervised fine-tuning was employed to train
using reinforcement learning based on human
feedback. Additionally, coaches received
written advice to assist them in writing their
submissions. In order to create a dialogue
format, they combined the InstructGPT
dataset with this new dataset. The flow
diagram is shown below.

13 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

apply categorization for this unique use

DATA SCIENCE case in order to assign each rule to a certain


slice. In the production environment, there

AT WORK
was a problem with the dataset's hydration.
Therefore, users were unable to access the
data from this dataset for reporting purposes.
Although initially not important, the reports
were required for the quarter-end reporting.

He used Python as his tool and


multi-label KNN classification as his method
of classification. Using the binary classification
technique was insufficient for him because
he had five slices that he wanted to tag
with the rules. He therefore decided to use
many labels for categorization. In addition,
he desired that the rules in a given slice be
functionally organised according to their
usage, effectiveness, and scope. He came
SHRIYAM up with KNN categorization by taking these
PGP-DSBA PROGRAM ALUMNUS distinct use cases into consideration.
archushukla123@gmail.com
BK8EHPNQZI
Since Shriyam has spent his entire career Here, the procedure was as follows:
working in the data domain, selecting this He had to generate a dataset with all the
program was more of an upskill decision and various rules' metadata, establish dummy
an attempt to transition from his current Data variables for all string-valued characteristics,
Engineering role into the Data Science role. and then apply the algorithm to this dataset.
Hence he chose to enrol in the PGP-DSBA Performance was assessed after each
Program offered by Great Learning. Shriyam classification output.
is currently employed as a Data Engineer II
The key difficulty was choosing the variables
and works with his stakeholders to ensure
properly such that the KNN algorithm would
proper data is available for their reporting
yield a classification that was based on
requirements. This work includes setting up
business logic rather than just chance.
a data pipeline, maintaining data quality, and
ensuring timely availability of data. The KNN algorithm's new slices, however,
were quite effective and decreased the total
He has encountered a few use cases while
processing time from 70 minutes to just under
working as a Data Engineer where he might
25 minutes. This enhanced the dataset's
employ Data Science principles. Python and
performance by 64% and increased its
clustering, however, have been crucial in
scalability.
assisting him in designing effective pipelines
to guarantee the timely availability of data He had the opportunity to try and solve a
and also improve data quality. Adding data practical problem using the knowledge he had
slices to one of the datasets and arranging learned in the program modules through this
the rules according to some kind of business exercise.
rules were two examples. He then had to

14 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

AI AT
that are similar across all failing test logs,
he decided to use cosine similarity. He was
then able to determine the particular failures

WORK along the regression. In addition, he utilised


K-Means Clustering to put all test cases with
comparable faults into one group so that
designers could only troubleshoot one test
log among a group of test cases that had the
same failure.

For Cosine Similarity:

1. He collected all the error logs and formed a


dictionary of unique words

2. Then for each error log in each test case,


formed the word frequency vector

3. Once frequency vectors are formed,


run a cosine similarity across all pairs of
KIRAN KUMAR MURALIDHARAN messages and form a Cosine similarity
PGP-AIML ONLINE COURSE Score Matrix
archushukla123@gmail.com
BK8EHPNQZI 4. On the above Cosine similarity score
It is common practice to regularly run 1000s Matrix, did another Cosine Similarity to
of tests in a regression suite in System On eliminate all the messages which correlate
Chip (SOC) Verification. This is done to with another message to the same degree
check the design's stability over numerous
regression runs using different random seeds. 5. Finally had a handful of messages with
The failure logs of such regression results are Cosine similarity less than say 99%
frequently difficult to analyse when the design
This handful of messages were the unique
is in an early stage. There may be numerous
Error signatures to be debugged
failed tests, and there may be numerous
test-specific failure locations. The failures
As a next step, for each test case, he created
would actually only be the result of a small
a ‘Test Case-Unique Error’ Message Matrix.
number of widespread design mistakes. In
For each Test Case, it will say which error
order to make the job of his team to analyse
message is present in it. He applied K-means
the problem easier, Kiran set out to discover
Clustering on this to get a picture of which
a technique to apply some machine learning
test cases have similar error signatures. Finally,
algorithm to analyse all failures across all test
recommend only the handful of test cases that
cases and produce a distinct error report.
cover all unique errors in that regression run.
The language of the error messages in the
test logs was typically comparable (but With the above solution, he could reduce the
not identical). However, there were certain problem space from 1000s of test cases * 10s
contextual variations. For instance, each of errors to a handful of test cases * handful of
communication had a unique address or errors and also reduce man-hours to 2-3 hours
set of data. In order to identify statements from 2-3 days.

15 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

MENTOR
Q2. How did you decide you want to be a
Data Scientist?

SPEAKS
I was drawn to the area of Data Science
because of its ability to generate insights
and value from vast volumes of data. This is
how my personal preferences or decisions to
become a data scientist may be described.
To analyse data using a range of tools and
methods, spot patterns and trends, and
create predictive models that can guide
corporate choices and spur innovation. I was
intrigued by the challenge of using data to
solve complicated problems as well as the
possibility of finding well-paying employment
in this quickly expanding industry.

Strong backgrounds in statistics,


programming, and machine learning methods,
In this edition, we will hear about our as well as an enthusiasm for using data to
mentor Jyant Mahara’s journey of draw conclusions and address practical issues,
becoming a Data Science industry expert.
archushukla123@gmail.com are characteristics I truly value as I start along
BK8EHPNQZI the path to becoming a data scientist.

Q1. Describe your current role. Q3. What preparations did you do to
I am currently leading a team at Zscaler to achieve your goal?
develop two models: an attrition model and I took the following actions to develop my
a chatbot using AWS LEX. I have a strong data science skills:
track record in addressing various challenges,
1. Education: I obtained my PGP-AIML
including Regression model for promotion
Online Course certification from Great
optimisation, Seaborn and Matplotlib for
Learning
exhaustive EDA, CNN, RESNET50 for image
attribute tagging, NLP for data extraction,
2. Technical talents: In addition to soft
Movie recommendation engine using Lightfm
talents, I was proficient in Python, R,
Neural network model, Price Optimization
and SQL programming languages.
Model, and Credit Scoring Model using bank
understanding of visualization using
and retail data.
programs like Tableau, Power Bi, and
Superset.

3. Analytical Skills: The ability to analyze


vast volumes of data and derive insightful
conclusions from it. I also focused on my
problem-solving abilities and enhanced my
capacity for clear communication of my
results.

16 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

4. Industry Knowledge: Data scientists need 2. Acquire programming skills: Python, R, and
to understand the business issues they are SQL are a few of the computer languages
attempting to solve as well as the industry that data scientists frequently employ.
in which they work. Learn at least one language well.

5. Continuous Learning: Because the area of 3. Practice on real-world issues: To get


data science is always changing, it's critical experience, try tackling issues like
for data scientists to stay current on the anticipating customer churn or examining
newest tools and methods. sentiment on social media. You can
develop your abilities by taking part in
6. In conclusion, acquiring the necessary online contests or contributing to
education, technical abilities, analytical open-source initiatives.
skills, business understanding, and ongoing
learning is necessary to become a data 4. Maintain your curiosity: Because the
scientist. subject of data science is continuously
changing, it's critical to maintain your
Q4. How did you get your first job and curiosity and keep studying. To keep
describe your journey (difficulties that you current on the newest trends
faced and how did you overcome)?
I struggled a lot to find my first work in the
archushukla123@gmail.com
BK8EHPNQZI
data science industry because I was new to
it and was mostly seeking a fresher's role. I
made the decision to focus on industry-related
initiatives that could help me better grasp
real-world issues because company rejections
were depressing me. I had the opportunity to
work on several projects at GL that gave me
insight into the workings of the
machine-learning industry. Additionally,
practicing different Kaggle puzzles was
helpful. After all of that, I was given the option
to work for a firm that produces AI-based
products, giving me the chance to hone my
talents even further.

Q5. Advice to DSBA learners?


A few pointers for people just getting started
with AIML/DSBA:

1. Solid Foundation: You may create a


solid foundation by developing a solid
foundation in math and statistics.

17 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

CROSSWORD PUZZLE

archushukla123@gmail.com
BK8EHPNQZI

ACROSS DOWN

6. The primary algorithm for performing 1. State-action value function.


gradient descent on neural networks.
2. The more common label in a class-
7. A class of algorithms for pattern analysis, imbalanced dataset.
whose best-known member is the suppor
vector machine (SVM). 3. An ensemble approach to finding the
decision tree that best fits the training data.
8. A computer program that plays the board
game Go. 4. Gauges whether a computer-based
synthesized voice can tell a joke with
10. A popular Python machine learning API. sufficient skill to cause people to laugh-

5. Abbreviation for generative adversarial


network.

9. _____­­is the process of transforming any


given key or a string of characters into
another value.

18 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.
APRIL 2023 EDITION

LEARNING BIRD CHIRPS


Procrastination makes easy things
hard and hard things harder.

THE EDITORIAL TEAM

archushukla123@gmail.com
BK8EHPNQZI

Mugdha Deepala Anamika Singhal

CREATIVE TEAM

Amit Gaonkar
(Design)

19 This file is meant for personal use by archushukla123@gmail.com only.


Sharing or publishing the contents in part or full is liable for legal action.

You might also like