Building with Intelligence: A Hands-On Practitioner's Guide to LLMs, Agents & AI Systems (The AI Masterclass Series)
Format:
Paperback
En stock
1.26 kg
Sí
Nuevo
Amazon
USA
- Every AI system that ships to real users was built by someone who had to figure it out under pressure. This book is for that person.Building with Intelligence is a complete technical practitioner's guide to building production-grade systems with large language models — from how transformers actually work to multi-agent architectures, from your first API call to a deployment that survives real traffic. Fourteen chapters. Fifty-plus production code labs. Every concept paired with working Python and TypeScript side by side.What you will build:Starting from first principles, you will construct every major component of a modern AI stack: a retrieval-augmented generation pipeline backed by pgvector, a ReAct agent with explicit reasoning traces, a persistent memory system with episodic recall, a three-agent pipeline with an independent critic, a complete evaluation harness with LLM-as-judge scoring, a browser automation agent using Playwright, a defense-in-depth guardrail stack, a semantic cache, a model router that dispatches queries to the right model tier, and a full observability layer with cost attribution by user and feature.The capstone chapter integrates every component into a single production-ready system.Structured in three parts:Part I — The Intelligence Layer covers how LLMs work at the token level, what embeddings represent, how to engineer prompts that work reliably, how to build RAG pipelines that cite their sources, and when and how to fine-tune with LoRA and QLoRA.Part II — Building Agents covers the complete agent stack: tool use and the Model Context Protocol, the ReAct and plan-execute architectures, memory beyond the context window, multi-agent coordination with trust boundaries, evaluation infrastructure, full-stack agent capabilities including code sandboxing and file system access.Part III — Shipping and Sustaining covers production safety and guardrails anchored by the Air Canada chatbot liability case, streaming with parallel guardrail checking, semantic caching, model routing, observability and cost attribution, and the practitioner's playbook of patterns and anti-patterns distilled from real deployments.Every chapter opens with a real case study: the Bing Sydney incident, the InstructGPT breakthrough, Cursor's code editor, Devin's mixed reception, the Scale AI origin story, Anthropic's computer use release, the Air Canada tribunal ruling, and the startup that received a $400,000 API bill after a viral launch.This is not a survey of what AI can do. It is a builder's guide to doing it — reliably, safely, and at production scale.
IMPORT EASILY
By purchasing this product you can deduct VAT with your RUT number
★ APPLIES FRANCHISE