Agent Evaluation & Training

​Agents are only as good as the way we test them and the data they learn from. This evening we’ll bring together researchers, engineers, and builders who work on how AI agents are evaluated and trained, and talk through what actually works in practice.

​We’ll cover how persona-based simulated users can surface how different segments behave when they use an agent, where current evaluation metrics fall short, and how evaluation signals can feed back into model training. There will be time for open discussion, so bring your hardest evaluation problems and your own experience with what has and hasn’t worked.

​Food and drinks will be provided. Space is limited, so please RSVP below. We’ll confirm your spot by email.