The third volume of A Practitioner's Journey. Twenty-one chapters cover the modern AI stack end to end - large language models, retrieval-augmented generation, RLHF alignment, autonomous agents, and the LLM-specific infrastructure that surrounds them.
This is not another "build a ChatGPT clone" book. It is a working engineer's guide to the techniques and engineering decisions behind production AI systems. You will start with NLP foundations and tokenization, move through transformer architectures and LLMs, build production-grade RAG pipelines with vector stores and reranking, train reward models and align LLMs with RLHF and DPO, and finish with autonomous agents, the Model Context Protocol (MCP), multi-agent coordination, and multi-modal models.
You will learn to:
Build and deploy production RAG with vector stores, reranking, and citation groundingFine-tune and align LLMs with RLHF, DPO, and Constitutional AIDesign autonomous agent loops with safe tool use and approval gatesCoordinate multiple agents through the Model Context Protocol (MCP)Right-size LLM infrastructure across Bedrock, Vertex, Groq, and on-premDistill large models into deployable, edge-ready footprintsCompanion volumes: Books 1 and 2 of A Practitioner's Journey cover classical ML, deep learning, and the production engineering stack - recommended as prerequisites if you are new to the field, optional if you already work in ML/AI.