Train, evaluate, debug, and deploy ImageNet classifiers with runnable PyTorch code—and understand what each benchmark number actually means.
What if learning from data is not enough to make AI truly intelligent? Neural networks are remarkably good at discovering patterns, while symbolic AI offers a way to represent knowledge, rules, and logical relationships. Neuro-Symbolic AI explores how these two approaches can work together. This book takes you beyond the terminology to explore the ideas, architectures, and design approaches behind this emerging field. Discover how learning and reasoning complement each other, why combining them is challenging, and what this could mean for the future of AI. Understand the ideas first. Then explore how to build them.
Laid off, or sure you are next? Treat the severance like a seed round: protect the runway, keep the job search alive, and test one small product in writing before the money runs out.
ИИ может написать за вас почти всё — даже оправдание, почему это было честно. Через истории, научные исследования и неудобные вопросы Тимур Иоселиани исследует границу между полезной помощью и нечестным преимуществом.
Build governed, secure, and observable Copilot Studio agents from foundations through production. Visual explanations and examples cover orchestration, knowledge, tools, identity, testing, and operations.
Turn AI agent governance into enforceable architecture and operating controls. Visual explanations and examples cover authority, security, accountability, evaluation, and production assurance.
Design controlled, production-ready AI agents using practical patterns for tool use, retrieval, memory, planning, approvals, security, and operations.
Un parcours continu à travers les six phases du cycle de vie CPMAI, porté par une étude de cas unique de la première question métier jusqu'à l'exploitation. L'annexe A relie les 37 tâches des Grandes lignes du contenu de l'examen au chapitre qui enseigne chacune, de sorte qu'un point faible renvoie à un chapitre plutôt qu'à une recherche. Écrit pour les chefs de projet qui doivent livrer des projets IA sans devenir ingénieurs en IA.
Your AI agent can follow the wrong instruction, expose customer data, issue duplicate payments or execute an action nobody authorised.A prompt won't reliably stop it.This hands-on Python engineering handbook shows you how to build the runtime controls that do: least-privilege permissions, policy enforcement, human approval, kill switches, tamper-evident audit trails and safe recovery.Build one runnable reference agent, attack it with seven scenarios, and see what happened when two real language models were put through the same tests.Stop trusting agents to behave. Build systems that control what they can do.
A practical guide to testing and defending AI applications, from prompt injection and RAG to MCP tools, autonomous agents, and incident response.
Ship production AI on Claude Opus 5.5: tool use, MCP, retrieval, evals, prompt-injection defense, cost tuning, and a full FastAPI service.
Eight complete Laravel AI applications, built end to end with every trade-off argued, from a SQL agent that can’t be talked into leaking data to a voice assistant on a latency budget. Then use models to write, test, and review your own code without losing control of quality.
Может ли искусственный интеллект развиваться благодаря собственной истории? Книга о том, как действия и их последствия могут менять интеллектуальную организацию системы — и как проверить это экспериментально.
Prompt injection has no patch — only architecture. Threat-model LLM apps, then defend them with provenance, least-privilege tools, output gateways, and CI red-team gates.
Build reliable RAG and Agentic RAG systems from retrieval fundamentals to production engineering and controlled agent operation. Visual explanations and worked examples cover evaluation, security, grounded answers, observability, and troubleshooting.