Books

Learn how modern LLM inference engines work by building one from scratch in Rust. From transformers and tokenization to KV caching, quantization, batching, and GPU optimization, this book combines theory, hands-on code, and performance engineering to help you create fast, production-ready AI systems.

LLMs are too big for single GPUs, but quantization fixes that. This book cuts through the hype to show you how to actually compress models using GPTQ, AWQ, GGUF, and NF4 without losing quality. You get real benchmarks, working code, and a clear way to pick the right tool for your hardware. Stop guessing and start deploying efficient models today.

Game producers already have most of what AI collaboration requires. They have been managing context, decomposing work, and evaluating output for years. This book is the map from where they are to where the work is going.

Systems: How AI Scales reveals how production AI handles thousands of requests, shared models, queues, state, failures, security, evaluation, cost, and continuous change. Through intuitive stories, practical architectures, and accessible mathematics, discover what it takes to transform an AI capability into a service people can trust.Volume V of The AI Systems Series.

A model can recommend an action. But what gives it permission to act—and how does it verify what happened? Agent explores tools, authority, planning, execution, recovery, security, and oversight. It begins where Reasoning ends and leads to Systems, where one successful action must become a dependable service.

A fluent answer is not necessarily a reasoned answer. Through stories, examples, visual models, and real mathematics, Reasoning reveals how AI represents problems, searches possibilities, and verifies conclusions. It begins where Context ends—and stops at one crucial boundary: a decision is not an action. That is where Agent begins.

The known surrounds us, but only some of it becomes context. The unknown is what we ask AI to resolve. How does a fluent model know what matters—to this user, from this source, at this moment? Context follows one ambiguous instruction through retrieval, RAG, memory, graphs, MCP, provenance, security, and evaluation—showing how AI drives the known to derive the unknown.

Your company's answers are already written down. This book builds the machine that finds them.

AI in software development is about much more than generating code.Beyond Code Generation explores how AI can support the entire software development lifecycle—from understanding requirements and designing systems to development, testing, code review, deployment, and operations.

“If real users rarely stop at a single prompt, then testing shouldn’t either.” 

Make your Python AI code up to 100× faster using NumPy vectorization, Numba, parallel execution, and GPU acceleration. Learn through practical benchmarks and real-world optimization examples.

Prompt injection is the SQL injection of the AI era — and most LLM apps are wide open. This hands-on field manual shows application security engineers how to attack and defend LLM systems: direct, indirect, multimodal, and agentic injection, with working code, labs, and ASR benchmarks you can run in CI. Break it, then build the defense-in-depth that holds in production.

Move beyond the API. Dismantle the AI black box and build generative engines from scratch with pure Python and NumPy. Master the profound geometric principles and applied mathematics driving LLMs and Transformers. Transform from a mere consumer into an elite AI innovator by writing the core mathematical architecture yourself—no shortcuts, no frameworks, just pure engineering excellence.