Machine Learning on the Ground: Design and Operations of Real-World ML Applications

What is this book?

One of the most peculiar things about Machine Learning is the dichotomy between academic ML research and industrial ML production. Taking a glance at the flood of ML papers on arXiv, it seems that a new breakthrough is happening almost every week. Yet, how exactly we use these algorithms as part of an ML production system, which not only includes the ML model, but also a way to turn the model scores into meaningful business actions and ensure that these actions are the rights ones, is much less documented.

When I first started writing about Machine Learning and Data Science back in 2019, my goal has been to close this academic/industrial gap. Since then, I’ve published dozens of articles on the subject, and my audience has grown to over 1K followers, including ML students as well as ML practitioners. I published all of my writing on the online publishing platform Medium, and this book is a curated selection of 13 of my best pieces. I created this book:

• for readers who do not wish to purchase a Medium membership (which is required to read all of my articles),
• for readers who prefer to read in pdf format, which can be printed and marked up,
• for readers who want to have a permanent, curated selection of my best work, and lastly
• for readers who want to support me as a writer (I pay Gumroad 30% of each sale for promoting this book).
  

This book is organized in 4 themes,
(1) ML system design - how ML systems are designed, with some concrete examples,
(2) ML operations - how ML systems are operated, including topics such as monitoring and A/B testing,
(3) Advanced ML topics - class imbalance, language modeling with BERT, and reinforcement learning, and
(4) Beyond ML - thoughts that go beyond current ML practice.

About the author

I've been working on ML research and applications since 2018, first as a postdoctoral researcher at Argonne National Lab, then as data scientist at JP Morgan Chase, then as Applied Scientist at Amazon, and then as ML Engineer at Meta. My background is in Physics.

What you'll get

A single pdf file with 50 pages packed with useful ML insights from my best articles. Table of contents:

1. ML system design

• Learning to rank: A primer
• People You May Know: Behind the Algorithms That Bring Users Together
• The four maturity levels of ML production systems

2. ML operations

• The Joy of A/B Testing: Theory, Practice, and Pitfalls
• The Joy of A/B Testing, Part II: Advanced Topics
• Deploying Your Machine Learning Model Is Just the Beginning
• Is My Model Really Better? Why ML models that look good on paper are not guaranteed to work well in production

3. Advanced ML topics

• Class Imbalance in Machine Learning Problems: A Practical Guide
• What exactly happens when we fine-tune BERT?
• Reinforcement Learning: Machines that learn by doing

4. Beyond Machine Learning

• Algorithms are not enough
• Machine Learning: Science or Alchemy?
• The limits of knowledge