A 20-chapter textbook takes AI engineers from RNNs to agentic systems
Foundation Model Engineering: From Theory to Production
Foundation Model Engineering is a free online textbook by Seongeun So that traces how modern foundation models actually work, from the RNN-to-Transformer shift and MoE scaling through pre-training, RLHF, RAG, inference optimization, and agentic AI. It targets AI engineers who want to move past surface-level API usage, pairing conceptual explanations with PyTorch examples, quizzes, and interactive visualizers. The book frames evaluation and alignment as product problems, not just research topics, and is maintained as a living document open to contributions.
The payoff is not just more terminology. It is better engineering judgment.