Skip to content

Running this workshop

Thanks for running this workshop with your community! This section has everything you need: timing plans, a checklist, notes for each module and promo copy.

Attendees build an AI agent from scratch in Python, without a framework, for a made-up conference called the Lumen AI Summit. They then write evals to measure it, attack it in a capture-the-flag game, and defend it with five layers of security. Everything runs on attendees’ laptops with a free open model through Ollama, or on any OpenAI-compatible provider. No accounts, API keys or credits are required.

  • Developers new to AI. Some Python experience is enough. No AI background needed.
  • Developers who already use AI frameworks. They will see what’s under the hood and pick up evals and security, which most tutorials skip.
  • Not ideal for complete beginners to programming. For them, the facilitator can live-code Modules 1 to 3 while the room follows along.

This workshop is vendor neutral and framework free on purpose. It pairs well with workshops on MCP, LangChain and the Microsoft Foundry Toolkit: run this one first to teach the fundamentals, then one of those to go deeper on a specific stack.

Facilitator One person who has run through all modules at least once. See the facilitator notes.
Helpers Ideally one per 10 to 15 attendees, mostly for setup problems in the first half hour
Room Tables, power sockets for everyone, a projector
Network Wifi is only needed for setup. Once the model is downloaded, everything works offline.
Attendees A laptop with 8 GB of RAM or more and Python 3.10 or newer

The workshop is modular. Run all of it in a day, or pick parts for a shorter event. See Formats and timing.

The site is designed to be projected. Open each module page and walk through it with the room. The diagrams, tables and code blocks work as slides. Use the browser zoom (Ctrl or Cmd and +) so the back row can read it.

Translate it, swap the example data for a local event, or add a module. It’s all Markdown. See Contributing, and please send improvements back so every chapter benefits.

Share photos and highlights with your community, and send feedback or suggestions through GitHub issues. Feedback from organizers is how this workshop gets better.