TARTU workshop: AI agents in practice

The workshop introduces AI agents and how they differ from a regular chat with a large language model. For example, an agent can stay on a task for longer and turn ideas into prototypes.

In the first half you learn how a large language model works and when a system can be called an agent. We distinguish a planning workflow from a building workflow and look at the role of tools and skills. We discuss why a vague instruction gives a mediocre result, how to check the result, and what risks relate to data, copyright and security.

Then participants make a plan for their own idea or a given task (goal, input, output, quality criteria, risks), build a prototype with an agent and test it.

After a break, participants present their prototypes and describe the agent’s mistakes and the points where human control is needed. Everyone writes up their result as a digital learning bite. There is a Q&A on job-market skills and the safe use of agents, and the workshop ends with a feedback round.

No programming experience is needed, but curiosity and willingness to experiment are welcome. You need a laptop and a GitHub account, set up according to the guide sent before the workshop.

Tools used: OpenCode 1.19.4 (open-source agent software), the GitHub Codespaces cloud environment and OpenCode Zen free models. Everything runs in the browser in the cloud: nothing is installed on your computer, and the agent cannot reach your files. Because the free tools and models are used, text you enter reaches the model provider, so do not enter personal or other sensitive data.

The workshop is led by Heili Aavola, a junior researcher in machine learning at the Institute of Computer Science, University of Tartu. Her PhD work deals with managing uncertainty in hybrid AI systems, that is, whether an AI system can recognise when it may be wrong. Together with Martin Leissoo she built an AI agent called Hermes at the University of Tartu, which runs on the university’s own hardware. She holds a bachelor’s degree in chemistry and a master’s degree in data science.

The event is free.