Learn to build and secure AI.
Hands-on labs, full-length practice exams, and practitioner articles for working with Claude, from your first well-formed prompt to a production agent architecture that holds up under review.
AI is being adopted faster than it is being learned.
Most people using AI tools at work picked them up on their own. They learned what worked by trial and error, and nobody showed them where the sharp edges are: the prompt that leaks data, the agent with more permissions than it needs, the output that reads as confident and is simply wrong.
The material here is built to close that gap. Every track is organized around a published certification blueprint, every lab is something you do rather than something you read, and every practice question is written so that you cannot pass it by guessing.
Security is not a separate module. It runs through all of it, because that is where it belongs.
Pick your path
For anyone who uses Claude to get work done. No code at all. Eight guided labs in your own account covering prompting, checking output, choosing a model, Projects, and using AI responsibly at work.
Start the track →For engineers shipping Claude applications. Nine code-first labs across the API, tools and MCP, streaming, caching, cost, evaluation, secrets handling, and prompt injection defense.
Start the track →For engineers designing production systems. Twelve modules on agentic loops, the Agent SDK, MCP, Claude Code, and reliability, ending in a capstone orchestrator project, plus an advanced agent-building track.
Start the track →For people who own Claude systems end to end. Eleven design-studio labs on solution design, RAG, observability, evaluation, governance, and explaining trade-offs to stakeholders, closing with an architecture review.
Start the track →Not sure where to start? The certification catalog compares all four tracks side by side.
From the articles
What prompt injection is, why current defenses fall short, and what teams handling sensitive data should do about it.
Read →MCP servers bridge models to production systems. Where real deployments are failing, a working threat model, and what to build before production.
Read →Each risk in the list, translated from a category name into what it looks like in a real deployment and what you can do about it.
Read →Why static scopes break down for autonomous agents, and how to design just-in-time, task-based permissions with an audit trail that holds up.
Read →Open source
A spec-driven delivery pipeline for Claude Code: specification gates, test-driven builds, dual-phase QA, and requirement traceability. A working example of the discipline the Architect tracks teach, released as a plugin you can read and run.
View on GitHub →Written by practitioners who teach.
Everything here is written by people who build and secure AI systems for a living, and who also teach: in a university engineering program, and in a practitioner meetup for people building with Claude Code.
That combination shapes the material. The examples come from real systems, and the explanations are written for someone learning it for the first time.
Start with the track that fits your role.
Every track has labs you can do today and a practice exam to check yourself against.
Browse the courses →