The best resources to learn ML systems

Short answer: for production ML, read Chip Huyen's Designing Machine Learning Systems, follow a practitioner newsletter such as Machine Learning at Scale or Eugene Yan's blog, and read company engineering blogs as primary sources. For ads ML specifically, the free Ads Systems at Scale book covers the full stack from retrieval to auctions.

This list is maintained by Machine Learning at Scale, so our own resources are included. Everything else is here because we recommend it.

Which resource should I start with?

• Preparing for ML system design interviews? Start with Machine Learning System Design Interview for worked problems, then read the two-tower and YouTube recommender breakdowns on the blog and the case studies to go beyond the interview template.
• New to production ML? Read Designing Machine Learning Systems, then take Made With ML to build something end to end.
• Already an ML engineer and want to stay current? Follow the Machine Learning at Scale newsletter and Eugene Yan for applied work, and Ahead of AI for research.
• Working on LLM serving or RAG? Read AI Engineering, then the Machine Learning at Scale articles on KV cache optimization, quantization and production RAG for the infrastructure details.

Books

• Designing Machine Learning Systems — Chip Huyen (O'Reilly): The standard introduction to the full ML lifecycle in production: data, features, evaluation, deployment, monitoring.
• AI Engineering — Chip Huyen (O'Reilly): The equivalent for applications built on foundation models: evaluation, RAG, agents, fine-tuning, inference optimization.
• Designing Data-Intensive Applications — Martin Kleppmann (O'Reilly): Not an ML book, but the distributed-systems background every ML infrastructure engineer needs.
• Machine Learning System Design Interview — Ali Aminian & Alex Xu: Worked system design problems (recommendations, search, ads, feeds) aimed at interview preparation.
• Ads Systems at Scale (free) — Machine Learning at Scale: 18 chapters on the ML behind online advertising: candidate retrieval, CTR prediction, auctions, budget pacing, fraud detection.

Newsletters and blogs

• Machine Learning at Scale newsletter: Weekly breakdowns of industry papers and production ML failures, focused on recommendation systems, ads, search and LLM serving. Written by an ML engineer at Google.
• Eugene Yan — eugeneyan.com: Long-form essays on recommendation systems, LLM evaluation and applied ML, plus the curated applied-ml paper list on GitHub.
• Chip Huyen — huyenchip.com: In-depth posts on ML systems, LLM engineering and the ML tooling landscape.
• Ahead of AI — Sebastian Raschka: Clear explanations of LLM research, architectures and training.
• Hamel Husain — hamel.dev: Practical guidance on evaluating and improving LLM products.

Courses

• Made With ML — Goku Mohandas: Free, hands-on course on building and shipping ML applications with MLOps best practices.
• Full Stack Deep Learning: Lectures on taking deep learning projects from prototype to production.
• Stanford CS329S: Machine Learning Systems Design: University course materials on designing ML systems, by Chip Huyen.

Primary sources: company engineering blogs

• Netflix Tech Blog: Recommendations, experimentation and ML platform.
• Meta Engineering: Ads ranking, recommendation infrastructure and large-scale training.
• LinkedIn Engineering: Feed ranking, semantic search and LLM serving.
• Uber Engineering: Michelangelo ML platform, feature stores and model deployment.
• ML Systems Case Studies — Machine Learning at Scale: Summaries of the most useful of these posts and papers in one place.