Artificial Intelligence for Engineers: Transformers, LLMs, RAG, and AI Agents: from foundations to code - Couverture souple

Salemme, Luciano

Couverture souple

ISBN 13 : 9798161244357

Editeur : Independently published, 2026

Synopsis

What really happens between a prompt and a language model’s response?

Why can an LLM write code but get a multiplication wrong?

And what does it actually take to bring a system built on these models into production without unpleasant surprises in terms of cost, security, and regulatory compliance?

This book approaches these questions from an engineering perspective: first, understand the mechanisms, working through every calculation by hand down to the last digit; then, implement them in executable Python code. No magic, no black boxes: every component is taken apart, explained, and put back together.

PART I. FOUNDATIONS

The journey from neural networks to large language models: a neural network trained entirely on paper, with backpropagation verified digit by digit in PyTorch; BPE tokenization performed by hand; embeddings and the geometry of meaning; the attention mechanism reconstructed number by number; the complete Transformer architecture, culminating in a working mini-GPT built in just a few lines of code; and the process of turning a raw model into an assistant, covering instruction tuning, alignment with human preferences, quantization, and open models.

PART II. BUILDING WITH LLMs

Professional use of models through APIs: streaming, function calling, and structured output; prompt and context engineering as a discipline, including defenses against prompt injection; semantic search and vector databases; and a complete RAG pipeline, from chunking to evaluation, with verifiable citations.

PART III. AGENTS AND MULTI-AGENT SYSTEMS

The agent loop, with tools and guardrails; the Model Context Protocol (MCP) for connecting agents to enterprise systems; multi-agent orchestration patterns, the A2A protocol, and LangGraph, illustrated through a real-world B2B logistics case study that shares the actual design decisions and lessons learned.

PART IV. IN PRODUCTION

Architecture and cost management, gateways and graceful degradation; security; the GDPR and EU AI Act applied to LLM-based systems; observability and evaluation using golden datasets and LLM-as-a-judge techniques; and a final case study that brings together every component covered in the book to build a complete enterprise assistant: permission-aware RAG, agents with human approval, MCP servers, end-to-end request tracing, and a fully worked cost analysis.

WHO THIS BOOK IS FOR

Engineers, developers, and students who want more than just to use AI: they want to understand how it works and build robust production-ready systems on top of it. A basic knowledge of Python and first-year university mathematics is all you need. Everything else is developed step by step throughout the book.

Seventeen chapters with complete, executable code examples: a journey where theory explains the concepts and working code brings them to life.

Détails de l'édition

• Éditeur: Independently published
• Date d'édition: 2026
• Langue: anglais
• ISBN 13: 9798161244357
• Reliure: Broché
• Nombre de pages: 384
• Coordonnées du fabricant: Printed by Amazon on behalf of the author; https://www.amazon.fr/hz/contact-us; ; c/o Amazon Media EU S.à.r.l., 38 Avenue John F. Kennedy; Luxembourg; L-1855; Luxembourg