The Mechanics of LLMs: Theory, Architecture, and Practice for Engineers

Don’t Just Use LLMs. Master Them!Large language models (LLMs) — ChatGPT, Claude, LLaMA, and more — are everywhere. But truly understanding how they work is a rare skill. The Mechanics of LLMs brings French engineering expertise directly to your hands, combining rigorous theory with practical, hands-on guidance.This book guides you step by step: from individual tokens to Transformer architectures, from training to fine-tuning, and all the way to agentic systems and advanced inference. It’s a practical and theoretical handbook designed for engineering students, developers, and self-learners who want to go beyond “black-box” usage.Inside, you’ll discover:Python scripts and practical examples to test every concept yourself.The secrets of modern architectures: Transformers, multi-head attention, RoPE, Flash Attention.Advanced adaptation techniques: LoRA, QLoRA, fine-tuning, RLHF.Advanced applications and agentic systems: RAG, optimized inference.Appendices, glossary, and references to master modern frameworks and vocabulary for years to come.Who this book is for:AI and computer science studentsEngineering studentsDevelopers and engineers who want to go beyond the black boxSelf-learners determined to understand LLMs deeplyWhy this book stands out:One of the few comprehensive guides from France offering a full, technical, operational view of LLMs, far from oversimplification or hype. Designed with a strong pedagogical focus, it combines clear explanations with practical examples that anyone can follow.About the author:The author is an engineer with a background in French higher education, over 20 years of experience in IT leadership as a Chief Information Officer, and extensive experience in teaching. He combines practical engineering expertise with a focus on explaining complex concepts clearly and accessibly. Read more