Building LLM agents

Join us for a one-day hands-on course on building LLM agents. You start with your first call to a language model and end with a multi-agent system that you orchestrate yourself. Everything is built from scratch in plain Python, so you see what actually happens underneath tools like LangGraph, CrewAI and the OpenAI Agents SDK.

The course is taught live in Jupyter notebooks. The instructor goes through each notebook step by step while you run the same code, either on your own laptop or in an environment we provide. The models run on Sigma2’s AI inference service (ai-gateway.sigma2.no).

Topics include:

• Prompting and structured output 
• Tool calling and the agent loop 
• Memory and context management
• Guardrails
• Orchestration patterns such as pipelines, routers and planner-executor setups 
• What it takes to run agents in production

The course is for researchers, students and developers who know basic Python and want to understand how LLM agents work. You don’t need any prior experience with LLMs or agent frameworks.

What to bring:

A laptop with a web browser. You can use the environment we provide, or run the notebooks locally with Python 3.