Skip to content
MeetupOnlinePast event

Closing the Continuous Learning Loop: From Agent Traces to Agent Training

Event artwork for Closing the Continuous Learning Loop: From Agent Traces to Agent Training

This event has ended

This listing is kept as an archive. Explore upcoming events to find your next one.

See upcoming events in Amsterdam
About this event

What to expect

On 21 September 2026, learn how to turn production AI agent traces into simulations, rewards and model improvements at this online engineering meetup.

Hosted by the AI Performance Engineering Meetup (Amsterdam), this 90-minute virtual session examines a central challenge in production AI: turning evidence of agent failures into a repeatable process for improving the system.

From traces to training

The main talk, “Closing the Continuous Learning Loop: From Agent Traces to Agent Training,” will be presented by Rustem Feyzkhanov, Engineering Manager at Snorkel AI. Production traces can reveal where an AI agent breaks down, but identifying a failure is only the beginning. Rustem will explore a practical learning loop that moves from observed behaviour to:

  • Reproducible simulations based on production traces
  • Verifiable rewards for evaluating agent behaviour
  • Model improvement through fine-tuning or reinforcement learning
  • Decisions about whether model weights need to change at all

The session will also consider situations where changing the prompts, tools or agent harness may be more effective than retraining the underlying model. This makes the event particularly useful for anyone trying to distinguish model problems from broader agent-system and orchestration problems.

Format and audience

Chris Fregly and Antje Barth will open with introductions and updates from the meetup before the featured technical talk. The event takes place online via Zoom from 18:00 to 19:30 on Monday, 21 September 2026. It forms part of the group’s regular third-Monday meetup series, organised by the AI Performance Engineering community in Amsterdam.

The subject matter is especially relevant to AI and ML engineers, agent builders, platform teams, engineering leaders and practitioners responsible for evaluating or improving production agent systems. No formal prerequisites are listed, although the focus on traces, simulations, fine-tuning and reinforcement learning makes this a technically oriented session.…

Register for this event

Running an AI event in Amsterdam?

List it here, free.

Meetups, workshops and conferences go in front of thousands of builders after a quick review.

List your event