AI Engineering: Building Applications with Foundation Models

This book presents a collection of helpful ideas and suggestions to aid engineers in developing AI applications on top of LLM models. I’ve already recommended it to the AI engineers that I work with.

If you’re at all interested in building products using large language models, this book is definitely a must read. The focus on evaluation and observability throughout multiple chapters was also refreshing to read as it explored a topic that at the heart of success/failure for AI products but is mostly in its infancy.

What a fantastic book! A great resource for people who interested in AI and its inner workings.

Yes- this is a very informative book.

Great comprehensive book on AI engineering. This book simplifies the concepts and techniques of advanced AI development with practical applications across Generative AI

A dense read, but insightful. Nice work.

I'm only up to Chapter 4, and this book is fantastic! I'm coming from a ML/deep learning background, looking to level up with generative AI and LLM's, and I think this book is great! I was hesitant at first, there is so much to find for free - but this book is concisely pulling it together with many interesting details! If I had been bouncing around on the web instead of reading this book, I don't think I'd know 1/2 as much about these early chapter topics as I do now! The book is getting me excited about working in the field - building an AI based product, leveraging existing models, digging into foundation model training - and there's still 7 more chapters to go!

In academia, there is the concept of a "review article" -- it summarizes and organizes the major research findings into a framework that makes it easy to come up to speed on a topic. Frequently, the review articles themselves end up defining the area, and this is what Chip Huyen manages to achieve in this comprehensive book. The quality of the writing and diagams are uniformly high -- Chip uses simple language to great effect.I think of myself as being somewhat up to date, but I have learned something new every chapter and not just minor details. For example, I had missed the Deep Mind paper pointing to "self-delusion" as the reason for hallucinations. Chip provides a clear explanation and shows an example. This fundamentally affects my intuitive understanding of model errors.Of course, there's a danger with writing a review of a fast moving field. Just today, DeepSeek published an article showing that they can avoid SFT altogether and do just train a model on preferences, alphago-style. If this takes off, Chapter 7 will need a second edition.Strongly recommend this book. It's invaluable for anyone building applications using GenAI models.