Building AI Agents with LLMs

What you'll build

This is a hands-on, project-based course on building real AI agents with large language models. Instead of disconnected demos, you build one product — Aria, a business assistant for the fictional Northwind Co. — and grow it layer by layer: a well-prompted model, then a looping LangGraph agent, then tools, retrieval, a multi-agent team, and finally a secured, evaluated, deployed service.

The journey

You start with prompt engineering — the single biggest lever on reliability — and structured outputs your code can trust. Then you build your first agent loop from LangGraph primitives so you understand the machine before reaching for shortcuts. From there you add real tools and API connections, a full RAG pipeline with citations, and multi-agent coordination. The final sections take Aria to production: monitoring, cost control, security and guardrails, evaluation, and a gradual rollout to real users.

Projects & capstone

Every major section ends in a build: Aria the Task Agent, the Operations Assistant, the Knowledge Assistant, a Multi-Agent Research System, and a production-ready capstone that integrates every layer. A closing Projects & Resources section collects the full specs, a one-time environment setup, the modern agent stack, and a plain-language glossary.

Who this is for

Developers and technical builders who can write basic Python and want to go from "I've called an LLM API" to "I can design, evaluate, and ship an agentic system." Each section also ends with a short quiz to check your understanding before you move on.