Large Language Models That Work: A Practical Python Guide to Transformers, Embeddings, Semantic Search, RAG, Fine-Tuning, AI Agents, Evaluation, and P

By Liam Everly

Build language-model systems that work after the demo.

Large language models can generate impressive text in minutes. Turning that capability into a reliable product is a different problem. Retrieval can miss the decisive source. A prompt can break on the next edge case. An agent can call the wrong tool. A model upgrade can improve a benchmark while making your real workflow slower, costlier, or less accurate.

Large Language Models That Work is a practical, production-first guide to designing AI applications that are bounded, measurable, and maintainable. Using clear mental models, original diagrams, and focused Python examples, Liam Everly shows how transformers, embeddings, semantic search, retrieval-augmented generation, fine-tuning, tools, and agents fit into one dependable system.

Inside, you will learn how to:
Understand tokens, attention, decoding, embeddings, and context without unnecessary mystique
Build semantic search with lexical retrieval, dense vectors, metadata filters, fusion, and reranking
Create RAG pipelines that handle missing evidence, conflicting sources, citations, permissions, and corpus updates
Use structured outputs for extraction, classification, and decision-focused summarization
Choose between prompting, retrieval, tools, workflow redesign, LoRA, and QLoRA
Evaluate correctness, grounding, calibration, latency, cost, safety, and slice performance before release
Design tool-using workflows and agents with schemas, approvals, budgets, idempotency, and audit trails
Operate production systems with tracing, canary releases, rollback, privacy controls, and incident response

• Format: Paperback
• Language: English
• ISBN: B0HBP7D8KV
• ISBN13: 9798189306471
• Release Date: July 2026
• Publisher: Independently Published
• Length: 196 Pages
• Weight: 0.59 lbs.
• Dimensions: 0.4" x 6.0" x 9.0"