Becoming An AIAP Field Guide V3.1 PDF

Becoming an AI Apprentice
A Field Guide (Version 3.1)
12 - 18 months self-directed AI/ML learning journey

Laurence Liew
Najib Ninaba
Cheong Wei Yih
Maurice Manning
Kevin Oh
Weng Jianshu
Ng Kim Hock
Jeanne Choo
Ryzal Kamis
Daniel Ng
Basil Han
February 2020
Becoming An AI Apprentice Field Guide by AI Singapore is licensed under a Creative Commons Attribution-NonCommercial-
ShareAlike 4.0 International License.
© 2020 AI Singapore 2
Preface
The AI Apprenticeship Programme (AIAP)™ is a 9-month long programme run by AI Singapore
(AISG) in collaboration with the Infocomm Media Development Authority (IMDA) to develop a
pipeline of local AI engineers for the industry. It is a deep skilling programme and selection into
it is rigorous. Applicants are expected to already have a solid foundational level of skills and
knowledge in the field of AI and machine learning (ML), albeit with limited hands-on experience
with and exposure to real-world data and production environments.
This field guide serves several needs. It offers a structured 12-month learning pathway to attain
the requisite level for acceptance into AIAP™. New applicants can check themselves against
the requirements outlined before they submit their applications. Unsuccessful applicants can
also evaluate their skills set against them and be motivated to improve and try again. Finally, it
defines the minimum standard in order to become an AI Associate Engineer in AISG’s AI
Certification pathway.
- The Editor, Basil Han
Introduction : Who is an AI engineer?
The AI (or ML) engineer is a relatively recent specialisation with a skills set overlapping with
those of the data scientist and data engineer [1]. An engineer in general builds things to solve
real-world problems. An AI engineer therefore harnesses AI technologies to build AI systems
that solve real-world problems.
To be an AI engineer requires strong conceptual understanding of AI/ML algorithms as well as

the requisite software engineering skills to operationalise and optimise AI systems in production.
He or she makes intelligent use of the research of data scientists (or works in close
collaboration with them) and, together with data engineers, builds systems that solve some
business or social problem. During this process, the AI engineer builds an actual
implementation which transforms a problem in the real world into a quantitative form which can
be manipulated to deliver something of value to the organisation within ethical and regulatory
boundaries. As can be seen, the AI engineering specialisation demands a wide range of skills
and knowledge.
To better prepare AIAP™ aspirants for selection into the programme, AISG has charted out a
12-month self-learning roadmap. It is a curation of online courses and resources which provide
the essential foundational knowledge grouped into 9 learning modules. Many of these resources
are free to access. Refer to the chart on the following page.
[1] Data engineers vs data scientists
The following sections provide more details on each of the modules.
Python and SQL (month 1)
Code is what animates computers. An AI engineer must be able to write, execute and debug
code as it is the means to translate concepts into real-world actions.
While there are many programming languages in the world, Python is the pre-eminent language
in AI as it is used to build many machine learning libraries and frameworks like scikit-learn,
TensorFlow and PyTorch.
● https://docs.python.org/3/tutorial/index.html
● https://www.youtube.com/playlist?list=PLlrxD0HtieHhS8VzuMCfQD4uJ9yne1mE6
(Microsoft’s 44-video series)
Code sharing and demonstration are part and parcel of AI engineering. Jupyter notebooks are
the de facto standard for this purpose.
● https://jupyter-notebook-beginner-guide.readthedocs.io/en/latest/
Data is the raw material for AI and is often housed in databases.
SQL is a standard language for accessing and manipulating databases. An AI engineer must
know how to use it to run queries that retrieve the required data from databases.
● https://www.w3schools.com/sql/sql_intro.asp
Software Engineering (month 1 - 3)
Microsoft Azure provides online learning resources for modern software engineering. It has
been chosen for this field guide for practical reasons, but you can use your favourite public
cloud provider course.
S/N Course Hours
1 What is Git? 2
2 Welcome to Agile 2
3 Develop Windows 10 applications 5
4 Azure Fundamentals - Cloud Computing 9
5 Azure for the Data Engineer 2
6 Work with relational data in Azure 4
7 Work with NoSQL data in Azure Cosmos DB 6
8 Administer containers in Azure 4
9 Run Docker containers with Azure Container Instances 2
Expect to spend 3-4 hours a week on the above for the next 8-12 weeks (2-3 months)
Data Science and Machine Learning (Part 1 : months 2 - 4)
The Data Science Design Manual
An excellent introduction and foundation course in Data Science and Machine Learning by
Professor Steven Skiena who is the Distinguished Teaching Professor of Computer Science at
Stony Brook University. His course and book The Data Science Design Manual provides an
excellent introduction with interesting war stories. Additional resources including data sets for
projects and assignments can be found at the website of the book.
This is a MUST READ for anyone who is starting out.
With the kind permission of Prof Skiena, AISG

has mounted his lectures as a course module
here.
Table of Content :
1. What is Data Science
2. Mathematical Preliminaries
3. Data Munging
4. Scores and Ranking
5. Statistical Analysis
6. Visualizing Data
7. Mathematical Models
8. Linear Algebra
9. Linear and Logistic Regression
10. Distance and Network Methods
11. Machine Learning
12. Big Data
Expect to spend 12-16 weeks (3-4 months) to complete this section.
Data Science and Machine Learning (Part 2 : months 4 - 6)
An Introduction to Statistical Learning (ISL)
The free PDF can be downloaded here and its associated video series (15 hours) and slides
here.
Note that the book uses R.
The equivalent Python notebooks kindly contributed to the community by Jordi Warmenhoven
can be found here.
Table of Content :
1. Introduction
2. Statistical Learning
3. Linear Regression
4. Classification
5. Resampling Methods
6. Linear Model Selection and Regularization
7. Moving Beyond Linearity
8. Tree-based Methods
9. Support Vector Machines
10. Unsupervised Learning
Expect to spend 12-16 weeks (3-4 months) to complete this section.
Deep Learning : Intel AI Academy (months 6 - 8)
Intel has developed a set of structured courses with hands-on exercises which everyone can
follow. It is free. You just need a PC or laptop. No GPU is required to complete the hands-on
exercises.
The courses are targeted at graduate

level and are meant for software
developers, data scientists and
students. Each course is for a 1-
semester module in a typical
university and lasts between 8 - 12
weeks and requires 3 hours per week
of your time. It is self-paced, so you
can finish it faster if you are able to.
You can access the Intel Academy

here.
We recommend you should aim to finish the above within 2-3 months.
Azure Machine Learning Service (months 9 - 10)
It is also recommended that you become familiar with an AI service from a cloud provider like
Azure. From the Azure Machine Learning Service page:
Azure Machine Learning service provides SDKs and services to quickly prep data, train,
and deploy machine learning models. Improve productivity and costs with autoscaling
compute & pipelines. Use these capabilities with open-source Python frameworks, such
as PyTorch, TensorFlow, and scikit-learn. Get started with our quickstarts and tutorials.
Complete the following step-by-step tutorials :
● Create your first classification model with automated machine learning

● Predict automobile price with the designer (preview)
● Get started creating your first ML experiment with the Python SDK
● Train image classification models with MNIST data and scikit-learn using Azure
Machine Learning
● Use automated machine learning to predict taxi fares
● Run batch inference on large amounts of data by using Azure Machine Learning
Plan to spend 1 - 2 months experimenting with the above.
Data Engineering (months 10 - 11)
To build effective machine learning models, you will require lots of data and must know how to
manage them. The following will help prepare you for the data engineering and data
management side of things. Azure has been chosen for practical reasons but you can opt to use
your favourite public cloud provider courses.
S/N Course Hours
1 Store data in Azure 4
2 Implement a Data Streaming Solution with Azure Streaming Analytics 2
3 Implement a Data Warehouse with Azure SQL Data Warehouse 4
4 Large Scale Data Processing with Azure Data Lake Storage Gen2 3
5 Building Open Source Software (OSS) Analytical Solutions with Azure 4

HDInsight
6 Big Data Analytics using Spark (edX) - optional 120
We recommend you should aim to finish the above within 2 months.
AI Ethics and Governance (months 11 - 12)
It is important that AI be built and used ethically, fairly and responsibly. An AI engineer should
be aware of these principles.
Model AI Governance Framework
Singapore released the first edition of the Model AI Governance Framework in 2019 to provide
guidance on key ethical and governance issues when deploying AI solutions. The second and
latest edition was released on 21 January, 2020.
● https://www.imda.gov.sg/infocomm-media-landscape/SGDigital/tech-pillars/Artificial-
Intelligence
● https://www.pdpc.gov.sg/Resources/Model-AI-Gov
AISG is leveraging this framework as part of its engineering best practices for the 100E projects.
The real-world experience from the projects delivered will be used to refine the framework
further.
Data Science Ethics Course
Enroll in an online course to learn about issues of data privacy and algorithmic bias and
fairness.
● https://www.edx.org/course/data-science-ethics
Additional Readings
● Microsoft: AI Principles & Approach from Microsoft

● Microsoft: FATE: Fairness, Accountability, Transparency, and Ethics in AI
● Microsoft: Responsible bots: 10 guidelines for developers of conversational AI
● Microsoft: The Future Computed (eBook)
● Google: Responsible AI Practices
● DeepMind: Safety & Ethics
● Interpretable Machine Learning
● Fairness and machine learning
We recommend you should aim to finish the above within 2 months.
Do AI Projects (months 7 - 12)
Put learning to practice by building an actual AI project. This can be through joining a
competition platform or, even better, building your own real-world application.
Competitions
Kaggle is the most popular AI/ML competition in the industry, and has
served as a starting point for many practitioners. It also allows you to
discuss with, learn from, and benchmark against other aspiring and
experienced AI engineers and data scientists.
https://www.kaggle.com/competitions
Below is a snapshot of other popular competitive platforms.
AI Makerspace
AI Singapore’s very own Makerspace platform contains curated

tools and resources for you to get started on solving your own
real-world AI problems.
https://makerspace.aisingapore.org/do-ai/
Work with NGOs
DataKind brings together data scientists and local nonprofit

organisations who require technical help. You can do good
while honing your skills and meeting like-minded
enthusiasts in person.
https://www.datakind.org/chapters/datakind-sg
Plan to spend 4-6 months experimenting with the above.
Recommended Readings
To supplement the journey, the following materials are also recommended. Note that some of
the books do get updated regularly.
Learning Python
● Python Crash Course, A Hands-On, Project-Based Introduction to Programming by Eric
Matthes
● Learning Python by Mark Lutz
● Fluent Python: Clear, Concise, and Effective Programming by Luciano Ramalho
● Learn Python the Hard Way by Zed Shaw, New York: Addison-Wesley
(https://learnpythonthehardway.org/)
Python for Data Analysis

● Python for Data Analysis: Data Wrangling with Pandas, NumPy, and IPython by Wes
McKinney
● Pandas for Everyone: Python Data Analysis by D Chen D (2017). New York: Addison-
Wesley.
● Python Data Science Handbook: Essential Tools for Working with Data by Jake
VanderPlas
Machine Learning
● The Hundred-Page Machine Learning Book by Andriy Burkov
● Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts,
Tools, and Techniques to Build Intelligent Systems 2nd Edition by Aurelien Geron
● Deep Learning with Python by Francois Chollet
● Deep Learning Illustrated: A Visual, Interactive Guide to Artificial Intelligence by Jon
Krohn , Beyleveld Grant
● Deep Learning Book at http://www.deeplearningbook.org/ (Advanced)
AI (Beyond Machine Learning)

● Artificial Intelligence: A Modern Approach by Stuart Russell
General AI
● On Intelligence: How a New Understanding of the Brain Will Lead to the Creation of
Truly Intelligent Machines by Jeff Hawkins
● Weapons of Math Destruction by Cathy O’Neil
● Rise of the Robots: Technology and the Threat of a Jobless Future by Martin Ford
● Human Compatible by Stuart Russell
● Architects of Intelligence by Martin Ford
● Rebooting AI – Building Artificial Intelligence We Can Trust by Gary Marcus, Ernest
Davis
● Automating Inequality by Virgina Eubanks

Becoming An AIAP Field Guide V3.1 PDF

Uploaded by

Copyright:

Available Formats

Becoming An AIAP Field Guide V3.1 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Becoming An AIAP Field Guide V3.1 PDF

Uploaded by

Copyright:

Available Formats

Becoming an AI Apprentice

A Field Guide (Version 3.1)

12 - 18 months self-directed AI/ML learning journey

- The Editor, Basil Han

To be an AI engineer requires strong conceptual understanding of AI/ML algorithms as well as

[1] Data engineers vs data scientists

S/N Course Hours

3 Develop Windows 10 applications 5

4 Azure Fundamentals - Cloud Computing 9

5 Azure for the Data Engineer 2

6 Work with relational data in Azure 4

7 Work with NoSQL data in Azure Cosmos DB 6

8 Administer containers in Azure 4

9 Run Docker containers with Azure Container Instances 2

This is a MUST READ for anyone who is starting out.

With the kind permission of Prof Skiena, AISG

Expect to spend 12-16 weeks (3-4 months) to complete this section.

Note that the book uses R.

Expect to spend 12-16 weeks (3-4 months) to complete this section.

The courses are targeted at graduate

You can access the Intel Academy

Complete the following step-by-step tutorials :

● Create your first classification model with automated machine learning

Plan to spend 1 - 2 months experimenting with the above.

S/N Course Hours

1 Store data in Azure 4

2 Implement a Data Streaming Solution with Azure Streaming Analytics 2

3 Implement a Data Warehouse with Azure SQL Data Warehouse 4

5 Building Open Source Software (OSS) Analytical Solutions with Azure 4

6 Big Data Analytics using Spark (edX) - optional 120

We recommend you should aim to finish the above within 2 months.

Model AI Governance Framework

Data Science Ethics Course

● Microsoft: AI Principles & Approach from Microsoft

We recommend you should aim to finish the above within 2 months.

Below is a snapshot of other popular competitive platforms.

AI Singapore’s very own Makerspace platform contains curated

Work with NGOs

DataKind brings together data scientists and local nonprofit

Plan to spend 4-6 months experimenting with the above.

Python for Data Analysis

AI (Beyond Machine Learning)

You might also like