dmls-book

This repository is the companion to Designing Machine Learning Systems, a book by Chip Huyen published by O'Reilly in 2022. The book is about how to build machine learning systems that work reliably in real-world production settings, covering decisions from handling data and choosing metrics all the way through deployment, monitoring, and ongoing maintenance. It focuses on the overall design and operation of ML systems rather than on teaching ML algorithms themselves.

The repository does not contain code examples. Instead it collects supporting materials that go alongside the book: a PDF table of contents, chapter-by-chapter summaries, a curated list of MLOps tools, a list of additional learning resources, and a short review of basic ML concepts. These files give readers a reference they can come back to without flipping through the full book.

The book is aimed at engineers, data scientists, and ML practitioners who are building or scaling ML systems at companies. It is explicitly not a beginner tutorial. The scenarios it addresses include things like deploying a model that performed well in experiments but behaves unexpectedly in production, setting up shared infrastructure like feature stores and model registries across a team, and detecting and fixing model degradation over time. Business and technical leaders without deep ML backgrounds may also find parts of it useful, particularly the earlier and later chapters.

The book has been translated into more than ten languages, including Japanese, Korean, Chinese, Portuguese, Spanish, Russian, and others, and is available through Amazon, O'Reilly, and Kindle.

Reviews quoted in the README come from practitioners at Google, Slack, and the Made With ML project. The repository itself is open for contributions via GitHub issues and pull requests.