BOOK TITLE: Azure Machine Learning Engineering

Subtitle: ML Engineering with Azure Machine Learning Service

AUTHOR: Dennis Sawyers, Sina Fakhraee, Ph.D. and Balamurugan
Balakreshnan

ABOUT THE AUTHOR

In order for us to write your author bio we need a few details. Please remember that your answers should be
relevant to the book. Your bio helps sell the book so please only include relevant information:

Full name Dennis Sawyers

What is your job title? Senior Cloud Solution Architect

What is the name of the organisation you work for? Microsoft

What is your skillset (as relevant to the book)? Data Science

In which industry do you work? Technology

What University degree do you have? Master’s from Carnegie Mellon University

What professional qualifications do you have? Azure Certified

PLease detail your relevant work history Microsoft - Cloud Solution Architect

Ford - Data Scientist

Relevant projects projects you have worked on:

1 Predict Pizza Demand for Worldwide Chain

2 Predict Player Performance for NBA Team

3 Predict Machine Failure for US Border Patrol

Would you like to include your social media details Yes. LinkedIn. (99+) Dennis (デニス　 ソーヤーズ)
(optional)? Sawyers | LinkedIn
ABOUT THE AUTHOR
In order for us to write your author bio we need a few details. Please remember that your answers should be
relevant to the book. Your bio helps sell the book so please only include relevant information:

Full name Sina Fakhraee, Ph.D.

What is your job title? Enterprise Data Scientist and Senior Cloud Solution
Architect

What is the name of the organisation you work for? Microsoft

What is your skillset (as relevant to the book)? Machine Learning and Data Science

In which industry do you work? Technology

What University degree do you have? Ph.D. in Computer Science and Engineering

What professional qualifications do you have? Azure Certified

PLease detail your relevant work history Microsoft - Enterprise Data Scientist and Senior Cloud
Solution Architect

Ford - AI/ML Product Owner and Technical Lead

Relevant projects projects you have worked on:

1 Developed a demand forecasting model for a global

retail customer.

2 Developed a predictive model for a utility customer

to classify power-outage reports to ones caused by
wiredown or non-wiredown.

3 Developed an end-to-end ML solution on kubernetes

for data scientists to process data, train models and
deploy their model.

Would you like to include your social media details https://www.linkedin.com/in/sina-fakhraee-ph-d-

(optional)? 2798ba70/

ABOUT THE AUTHOR

In order for us to write your author bio we need a few details. Please remember that your answers should be
relevant to the book. Your bio helps sell the book so please only include relevant information:

Full name Balamurugan Balakreshnan

What is your job title? Principal Cloud Solution Architect

 What is the name of the organisation you work for? Microsoft

What is your skillset (as relevant to the book)? Data/AI Architect and Data Science

In which industry do you work? Multiple

What University degree do you have? BE - Engineering

What professional qualifications do you have? Azure Certified

PLease detail your relevant work history Microsoft - Cloud Solution Architect

Relevant projects projects you have worked on:

Would you like to include your social media details

(optional)?

Can you recommend a good technical reviewer for your book?

[INSERT HERE: Name & Contact details]

PART ONE: BACKGROUND RESEARCH

TARGET AUDIENCE
Describe your target reader: what you assume about their knowledge of the topic, related topics, and
technical topics generally; why they want to learn the technology; what will they want to do with it?

Answer the following:

Who is your audience?

1
Data Scientists and ML Engineers who want to learn Azure

2
They know how to train and score machine learning algorithms
 What is important to them?

1
While they know how to train models, they do not know how to productionize them.

2
Lack of ability to deliver end-to-end solutions. Lack of knowledge how to automate machine
learning solutions. Lack of knowledge as to how to do these things on the Azure Platform.

3
Automate full end-to-end machine learning solutions using Microsoft Azure

COMPETITIVE BOOK TITLES

What is unique about your book? You will need to look on Amazon at books that have been well-received
– what are the top three market leading books that your book will compete with? Examine the
description, table of contents and book reviews.

List the books here:

1 Machine Learning Engineering by Andriy Burkov

2 Data Science on AWS: Implementing End-to-End, Continuous AI and Machine Learning Pipelines

3 Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control

Please ensure that you have looked at the description, table of contents and book reviews for each of
these books.

PART TWO: BOOK OVERVIEW

OVERVIEW
The long description is the device we use to describe the book on Amazon. Writing it is fairly systematic. Please
answer the following questions using only one sentence.

TEMPLATE EXAMPLE: Your turn...

Hands-On Blockchain Development with Hyperledger

Explain / Introduce the Hyperledger Fabric and Hyperledger Composer enable Azure Machine Learning Service to train machine
tech organizations to create private, permissioned learning models and productionize workloads
blockchain networks.

Why would a Data Scientists often have trouble productionizing

developer want to workloads and this will teach them how to do it.
learn it?
Why should they buy Developers working with Blockchain will be able to
this book? put their knowledge to work with this practical guide.
The book provides a hands-on approach to
implementation and associated methodologies that
will have you up-and-running, and productive in no
time.
Product approach Complete with step-by-step explanations of essential
concepts, practical examples and self-assessment
questions, you will begin by exploring the blockchain
evolution, including an overview of relevant
blockchain technologies
Product Breakdown: In You'll learn how to configure Hyperledger Fabric on a
2 sentences, describe cloud platform, understand the architectural
the “journey” the book components of Hyperledger Fabric, and how they are
takes the reader on. configured to build private blockchain networks, and
Look at your section the applications that connect to them. You'll then
headings for help build a network and application from scratch, and
learn how to implement smart contracts in chaincode.
By the end of this book By the end of this book, you will be able to build and
you will... deploy your own decentralized applications using
Hyperledger, addressing the key pain points
encountered in the blockchain life cycle.
Anything else you
would like to add?
LEARNING OUTCOME - WHAT WILL THE READER LEARN AND DO?
Consider the competing books; in particular the description, table of contents and book reviews. Decide
what the key learning objectives will be for your book. List them below:
1 Train Machine Learning Models in Azure Machine Learning Service
2 Build end-to-end Machine Learning Pipelines
3 Host Machine Learning Models on Real-Time Scoring Endpoints
4 Explain Machine Learning Models and Mitigate Bias
5 Use and MLOps Framework to Productionize Models
Data Scientists working with Azure will be able to
put their knowledge to work with this practical
guide to Machine Learning Engineering. The book
provides a hands-on approach to implementation
and associated methodologies that will have you
up-and-running, and productive in no time.
Complete with step-by-step explanations of
essential concepts, practical examples and self-
assessment questions, you will begin by training a
model in Azure Machine Learning Service followed
by step-by-step instructions in productionizing
your model.
You’ll learn how to train, register, and
productionize machine learning models using
Azure Machine Learning Service. You’ll learn how
to score models in real-time and in batch, how to
explain models to earn business trust, how to
mitigate model bias and develop solutions using an
MLOps framework.
By the end of this book, you will be able to fully
build and productionize end-to-end machine
learning solutions using Azure Machine Learning
Service.
PART THREE: BOOK STRUCTURE
Using your overview, and learning outcomes now decide on the structure of your book? What are your
start and end points?

GENERAL STRUCTURE
Divide the book into approximately 3 parts. The learning outcomes you listed previously will help to
inform these. These “parts” are a group of chapters that work toward the same goal. Each part will
consist of 3-5 chapters. For example: A book on Building Machine Learning Systems with Python might
be split into 5 parts as follows: “The Basics”; “Book Learning”; “Numbers, Forecasts and
Recommendations”; “Sound and Vision” and finally, “Practical Matters”.

WRITE YOUR PART HEADINGS BELOW:

1 Training and Tuning Models with Azure Machine Learning Service

2 Deploying and Explaining Models in AMLS

3 Advanced Topics - Monitoring, MLOps, Deep Learning, Distributed Training and Reinforcement
Learning

CHAPTER OUTLINE
Each chapter should have a clear focus. Each chapter title should clearly state what aspect of the overall
topic the chapter deals with.. Continuing the example of Building Machine Learning Systems with Python
your section on “Book Learning” might be broken down into 4 chapters as follows: “Clustering – sorting
text into groups”, “Topic Modeling – creating non-exclusive groups”; “Logistic Regression – evaluating
text quality”; “Bayes Classification – sentiment analysis”. PLEASE NOTE: Chapter titles appear on Amazon

PART ONE: Training and Tuning Models with Azure Machine Learning Service

1 Introducing Azure Machine Learning Service

2 Working with Data in AMLS

3
Training Machine Learning Models in AMLS

4 Tuning your models with AMLS

5 Scripting Automated Machine Learning Jobs

PART TWO: Deploying and Explaining Models in AMLS

6 Deploying ML Models for Real-Time Inferencing

7 Deploying ML Models for Batch Scoring

8
Explaining ML Models with AMLS

9 Mitigating Bias with AMLS

PART THREE: Advanced Topics - Monitoring, MLOps, and Deep Learning

10
Monitoring Models with AMLS

11 Productionizing your Workload with MLOps

12 Using Deep Learning in AMLS

13 Using Distributed Training in AMLS

14 Using Reinforcement Learning in AMLS

PLEASE ADD OTHER PARTS/CHAPTERS ONLY IF YOUR TOPIC REQUIRES IT

PART FOUR: DETAILED OUTLINE

PART 1: Training and Tuning Models with Azure Machine Learning Service

Readers will learn how to use Azure Machine Learning Service to train and tune different types of models
in part 1, taking advantage of its unique job tracking capabilities.

CHAPTER 1: Introducing Azure Machine Learning Service

- 25 pages

DESCRIPTION:

This chapter will introduce the basic concepts of Azure Machine Learning Service (AMLS). They will use
Chapter 1 to create an AMLS Workspace, create a compute instance, and connect AMLS to VS Code for
further development in later chapters.

Level: Basic

Main Chapter Headings (3-5 main chapter headings)

1. Building your first AMLS Workspace

2. Navigating the AMLS Studio
3. Creating Compute for Writing Code
4. Developing within AMLS
5. Connecting AMLS to VSCode

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Build an AMLS Workspace

2. SKILL 2: Learn how to use the AMLS Studio
3. SKILL 3: Create a Compute Instance for Developing and Running Code
4. SKILL 4: Open a Jupyter Notebook and/or Jupyter Lab
5. SKILL 5: Connect AMLS to VS Code

CHAPTER 2: Working with Data in AMLS

- 15 pages

DESCRIPTION:

In Chapter 2, readers will learn how to work with data in AMLS. In particular, they will learn how to load
data, save their data as datasets, and use datasets in later development projects.
Level: Basic

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Connecting AMLS to Storage Accounts

2. HEADING 2: Creating AMLS Datasets
3. HEADING 3: Using AMLS Datasets

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Create AMLS Datastores

2. SKILL 2: Load data into AMLS and register it as a dataset
3. SKILL 3: Call previously created datasets in your code

CHAPTER 3: Training Machine Learning Models in AMLS

- 25 pages

DESCRIPTION:

In this chapter, readers will learn how to train machine learning models using AMLS experiments as well
as the code-free designer. Readers will be able to train jobs remotely and save models to the AMLS
model registry for later use.

Level: Intermediate

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Creating Compute for Remote Jobs

2. HEADING 2: Making an Environment for Containerization
3. HEADING 3: Writing your First Model Training Script
4. HEADING 4: Registering your Machine Learning Model
5. HEADING 5: Training Code-Free Models with the Designer

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Create a Compute Cluster for remote job training

2. SKILL 2: Create an Environment
3. SKILL 3: Create an AMLS experiment
4. SKILL 4: Register a machine learning model
5. SKILL 5: Create a model training job using the code-free designer
 CHAPTER 4: Tuning your models with AMLS
- 25 pages

DESCRIPTION:

In this chapter, readers will learn how to tune hyperparameters for their machine learning models using
AMLS Hyperdrive.

Level: Intermediate

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Deciding which Parameters to Tune

2. HEADING 2: Writing your Hyperdrive Script
3. HEADING 3: Configuring your Hyperdrive Settings
4. HEADING 4: Running your Hyperdrive Script
5. HEADING 5: Parsing your Hyperdrive Results

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Creating a parameter space over which to tune

2. SKILL 2: Write a script to tune hyperparameters in AMLS
3. SKILL 3: Configure Hyperdrive settings to achieve best performance
4. SKILL 4: Select a strategy to tune your Hyperparameters
5. SKILL 5: Access and Understanding Hyperparameter tuning charts

CHAPTER 5: Coding an Automated Machine Learning Job

- 25 pages

DESCRIPTION:

In this chapter, readers will learn how to script an AutoML job to automatically train a machine learning
model.

Level: Intermediate

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Understanding Microsoft AutoML

2. HEADING 2: Writing your AutoML Script
3. HEADING 3: Configuring your AutoML Settings
4. HEADING 4: Training an AutoML model
5. HEADING 5: Parsing your AutoML Results
Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Identify when to use AutoML

2. SKILL 2: Write a script to run AutoML
3. SKILL 3: Configure AutoML settings for best performance
4. SKILL 4: Train an AutoML model
5. SKILL 5: Access and parse AutoML results

PART 2: Deploying and Explaining Models in AMLS

In part 2, readers will learn how to deploy models in AMLS in both batch and real-time. Additionally, they
will learn how to explain machine learning models and mitigate bias using AMLS.

CHAPTER 6: Deploying ML Models for Real-Time Inferencing

- 25 pages

DESCRIPTION:

In this chapter, readers will learn how to deploy machine learning models to real-time scoring endpoints
hosted on Azure Container Instances.

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Understanding Real-Time Inferencing

2. HEADING 2: Writing a Real-Time Scoring Script
3. HEADING 3: Creating a Real-Time Scoring Endpoint
4. HEADING 4: Deciding a Service to Host your Endpoint
5. HEADING 5: Scoring New Data in Real-Time

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Identify when to implement real-time inferencing

2. SKILL 2: Write a script to score models in real time
3. SKILL 3: Create a REST Endpoint
4. SKILL 4: Host a REST Endpoint on ACI and AKS
5. SKILL 5: Using the REST Endpoint to score new models
 CHAPTER 7: Deploying ML Models for Batch Scoring
- 25 pages

DESCRIPTION:

In this chapter, readers will learn how to score models in a batch using AMLS machine learning pipelines.

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Identify when to use Batch Scoring in AMLS

2. HEADING 2: Writing an AML Pipeline for Model Training
3. HEADING 3: Writing an AML Pipeline for Model Inferencing
4. HEADING 4: Publishing and Using Pipelines
5. HEADING 5: Scheduling Pipelines

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Understand when to use AMLS Pipelines

2. SKILL 2: Write an AMLS Pipeline for model training
3. SKILL 3: Write an AMLS Pipeline for model scoring
4. SKILL 4: Publish an AMLS pipeline
5. SKILL 5: Schedule AMLS Pipelines

CHAPTER 8: Explaining ML Models with AMLS

- 20 pages

DESCRIPTION:

In this chapter, readers will learn how to explain their machine learning models using AMLS and Azure
Interpret.

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Understanding Model Explainability

2. HEADING 2: Adding Explainability to your Training Runs
3. HEADING 3: Interpreting Models using Visualizations
4. HEADING 4: Adding Explainability to your Scoring Runs

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Explain models using SHAP

 2. SKILL 2: Add explainability to training runs
3. SKILL 3: Use all of the different explainability charts
4. SKILL 4: Add explainability to your scoring runs

CHAPTER 9: Mitigating Bias with AMLS

- 20 pages

DESCRIPTION:

In this chapter, readers will learn how to mitigate model bias using AMLS

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Understanding Bias in Machine Learning

2. HEADING 2: Adding Bias Detection to your Training Runs
3. HEADING 3: Reducing Bias in your AMLS Models
4. HEADING 4: Exploring the Bias-Performance Tradeoff

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Identify when model bias occurs

2. SKILL 2: Add bias detection to model training runs
3. SKILL 3: Mitigate bias using Fairlearn
4. SKILL 4: Explain to business end users how bias mitigation affects performance

PART 3: Advanced Topics - Monitoring, MLOps, and Deep Learning

In this section, readers will learn how to monitor machine learning models, use MLOps practices while
deploying models and run deep learning jobs with Azure Machine Learning Service.

CHAPTER 10: Monitoring Models with AMLS

- 20 pages

DESCRIPTION:

[INSERT HERE: 50-100 words, describing the content of the chapter]

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

 1. HEADING 1: Detecting Data Drift
2. HEADING 2: Monitoring Model Runs for Failure
3. HEADING 3: Using Application Insights
4. HEADING 4: Setting up notifications with Event Grid

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Detect data drift in your ML models

2. SKILL 2: Monitor ML Models using AMLS
3. SKILL 3: Use Application Insights for Job and Compute Monitoring
4. SKILL 4: Use Event Grid to set up E-mail alerts

CHAPTER 11: Productionizing your Workload with MLOps

- 20 pages

DESCRIPTION:

In this section, readers will learn how to integrate AMLS jobs with Azure DevOps and Github to achieve an
MLOps solution.

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

5. HEADING 1: Understanding MLOps

6. HEADING 2: Saving your code into GitHub
7. HEADING 3: Integrating AMLS with Azure DevOps
8. HEADING 4: Create a DevOps Pipeline
9. HEADING 5: Following Best MLOps Practices

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

10. SKILL 1: Create an Azure DevOps account

11. SKILL 2: Create a Github Account
12. SKILL 3: Link AMLS to Azure DevOps
13. SKILL 4: Create a DevOps Pipeline to deploy your code
14. SKILL 5: Use MLOps to productionize your work
 CHAPTER 12: Using Deep Learning in AMLS
- 20 pages

DESCRIPTION:

In this section, readers will learn how to use deep learning jobs in AMLS using multiple frameworks
including Pytorch and Tensorflow

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Creating a Tensorflow model with AMLS

2. HEADING 2: Creating a Pytorch with AMLS
3. HEADING 3: Creating an NLP model with AMLS
4. HEADING 4: Using Deep Learning Models in Production

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Train a model using Tensorflow in AMLS

2. SKILL 2: Train a model using Pytorch in AMLS
3. SKILL 3: Train an NLP model in AMLS
4. SKILL 4: Productionize a Deep Learning Model using an MLOps approach

CHAPTER 13: Using Distributed Training in AMLS

- 20 pages

DESCRIPTION:

In this section, readers will learn how to perform distributed machine learning in AMLS using Dask
framework and Modin DataFrame.

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Introduction to Dask and Modin

2. HEADING 2: Using Dask for distributed data processing
3. HEADING 3: Using Dask and Modin for distributed model training

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Quick introduction to Dask distributed framework and Modin DataFrame for Machine
Learning and how to set them up in AMLS
2. SKILL 2: Create a distributed client using Dask for parallel data processing
3. SKILL 3: Train a distributed XGBoost model using Modin DataFrame
 CHAPTER 14: Using Reinforcement Learning in AMLS
- 20 pages

DESCRIPTION:

In this section, readers will learn how to use Reinforcement Learning in AMLS using RLlib.

Level: Advanced

Main Chapter Headings (3-5 main chapter headings)

1. HEADING 1: Introduction to Reinforcement Learning and RLlib framework

2. HEADING 2: Train an agent to navigate through a lava maze in the minecraft game
3. HEADING 3: Train a cartpole playing agent on AML compute instance

Skills learned: For each heading, insert what the reader will learn to DO in this chapter?

1. SKILL 1: Quick introduction to reinforcement learning and RLlib framework and how to set it up in
AMLS
2. SKILL 2: Train a reinforcement learning agent using Ray RLlib to navigate through a lava maze in
the minecraft game
3. SKILL 3: Train a cartpole playing agent using Ray RLlib