Skip to content
Cover of Building LLMs for Production by Louis-François Bouchard & Louie Peters

Building LLMs for Production

by Louis-François Bouchard & Louie Peters

A solid breadth-first map of the LLM stack — just don't expect the framework chapters to age well.

4.0 my rating
4.10 public avg
Status
Read · November 2025
Bought
November 6, 2025
For
Engineers shipping their first RAG or fine-tuning pipeline · Backend devs who want the whole LLM stack in one pass · Anyone evaluating where prompting stops and fine-tuning starts

The one-paragraph verdict

This is the rare LLM book that actually goes start to finish — prompting, retrieval, fine-tuning, evaluation, deployment — without pretending any single one is a silver bullet. The strength is breadth and sequencing: it makes the why-now decisions legible, especially when prompting runs out of road and RAG or fine-tuning becomes the right tool. The weakness is the flip side of that breadth. It leans hard on LangChain and LlamaIndex, so a chunk of the code is already dating, and the editing is uneven — some chapters are tight, others read like rushed first drafts. Treat the prose as the product and the snippets as disposable, and it earns its keep.

Who should read it

Engineers shipping their first retrieval or fine-tuning pipeline who want the full landscape in one pass, rather than stitching it together from forty blog posts. It’s also a decent gut-check for backend devs deciding where prompting stops paying off. Skip it if you’re already running LLMs in production — you’ll find little here you haven’t learned the hard way.

Where it earned its place

The retrieval chapters mapped almost directly onto the work in improving RAG query quality — the book’s framing of query rewriting and reranking as separate, measurable stages is exactly the decomposition I leaned on in Archives. Its insistence on treating cost and token budgets as a first-class design constraint also fed straight into model policy & token governance.

Skip it if…

You want deep, current framework code or you’re past the beginner stage. The concepts hold up; the LangChain-heavy implementations will not, and a more advanced reader will outpace the material quickly.

#llm#rag#fine-tuning#ai-ml#prompt-engineering