AI Engineering¶
For people who write Python. By the end of this course you can design a multi-agent system with an orchestration layer, build a working prototype with an AI coding assistant, and evaluate, trace and secure it.
There is a self-paced pre-work session, then eight classes of about three hours. You need Python
3.11 or later, an editor with an AI coding assistant, and an API key for any model provider
(Azure OpenAI, OpenAI or Anthropic). Every lab's tests run without a key, and the lab code is in
the labs/ folder of the course download.
What you will build¶
Each class builds one layer of this architecture. The AI map explains every layer.
Classes¶
- Pre-work: setupA working environment and your first API call.
- Foundations: thinking and modelsBreak the task down; estimate tokens and cost. Layer: models.
- Prompting and structured outputsA SitRep extractor that returns validated data. Layer: harness.
- Building with code and AI pair programmingA spec-first app with retries and cost tracking. Layer: interface.
- Context engineering and RAGHybrid search with citations across languages. Layer: context and data.
- Harness and toolsA tool loop with guardrails and a step budget. Layer: harness.
- Agents and orchestrationA multi-agent SitRep with an orchestrator. Layers: orchestration and agents.
- Evals, observability and reliabilityA golden test set, an LLM grader and traces. Cross-cutting.
- Security, governance and productionA red-team harness and the capstone. Cross-cutting.