Building Conversational AI with LLMs and Agents: From the mathematics of attention to production agent systems. A practitioner's guide for engineers, researchers, and students
English | May 13, 2026 | ASIN: B0H1MQH23D | 5372 Pages | Epub | 71.13 MB
Build Production AI Agents from First PrinciplesThe 39-chapter LLM textbook that takes you from PyTorch fundamentals to shipping autonomous agents. 800+ pages of theory, code, and production engineering, written for engineers, researchers, and students who actually need to ship working LLM systems.If you have spent the last year trying to learn LLMs from scattered blog posts, GitHub READMEs, and outdated tutorials, this book is for you. Every chapter answers a specific question someone shipping LLM-powered systems has actually had to answer: how attention works, how to fine-tune a model on your data, how to build an agent that uses tools without going off the rails, and how to evaluate the thing you just built.What you will learnFoundations (Chapters 0-5): PyTorch and ML essentials, NLP and text representation, tokenization, sequence models, the Transformer architecture from scratch, decoding strategies.Understanding LLMs (Chapters 6-9, 18): pre-training, scaling laws, the modern model landscape, reasoning and test-time compute, inference optimization, and mechanistic interpretability.Working with LLMs (Chapters 10-12): APIs, prompt engineering, hybrid ML+LLM architectures.Training and adaptation (Chapters 13-17): synthetic data generation, fine-tuning, PEFT (LoRA, QLoRA), distillation, model merging, alignment with RLHF and DPO.Retrieval and conversation (Chapters 19-21): embeddings, vector databases, semantic search, retrieval-augmented generation, conversational AI systems.Agentic AI (Chapters 22-26): agent foundations, tool use and function calling, multi-agent systems, specialized agents, agent safety in production.Multimodal and applications (Chapters 27-28): vision-language models and real-world LLM applications.Production (Chapters 29-31): evaluation, observability, monitoring, engineering for LLM systems at scale.Safety and strategy (Chapters 32-33): alignment, ethics, regulation, product strategy, and ROI.Frontiers (Chapters 34-35): emerging architectures, the broader impact of AI on society.From idea to product (Chapters 36-38): shipping, scaling, and steering LLM-powered products.What makes this book differentTwenty tailored reading pathways. Whether you're a RAG engineer, agent builder, ML-to-LLM transitioner, product leader, or instructor, the book has a curated path through the chapters most relevant to your goal.Course-ready. Nine syllabi are included for graduate, undergraduate, and specialist tracks. Instructor copies and slide decks are available for course adoption.22 framework appendices covering Hugging Face, LangChain, LangGraph, CrewAI, LlamaIndex, Semantic Kernel, DSPy, plus environment setup, distributed ML, inference serving, and Docker.Up to date for 2026. Reasoning models, test-time compute, modern PEFT, agentic protocols (MCP, A2A), and multimodal foundation models all covered.Every chapter has a hands-on lab. You learn by building, not just by reading.
Links are Interchangeable - Single Extraction
