cs249r_book

This is Harvard's open-source textbook on Machine Learning Systems — not how to train a model, but how to engineer the entire system around it: deployment, edge inference, privacy, MLOps, and production constraints. It's the curriculum behind CS249r, now being published by MIT Press.

If you're an indie hacker building ML-powered products and want to understand why your model works in a notebook but fails in production, this book fills the gap. Fast.ai teaches you to train models. Stanford CS229 teaches the math. This book teaches the engineering. There's nothing else quite like it in the open-source textbook space.

The catch: It's an academic textbook, so the writing can be dense. The hands-on labs use TinyTorch and Marimo notebooks — niche tools your team might not know. Volume I and II arrive Summer 2026, so the content is still evolving. And "ML Systems" is broad — some chapters will be deeply relevant to your work and others won't apply at all. Read selectively, not cover-to-cover.