Zeta Alpha — Improving Multi-Agent Workflows with Self-Optimization

about this room

As part of Amsterdam Tech Week, Zeta Alpha's Arthur Camara leads an interactive talk on building multi-agent AI systems that improve themselves. Instead of endlessly tweaking prompts by hand, the session explores how agents can examine their own execution traces, reason about past failures, and refine their behaviour automatically using techniques like GEPA and TextGrad - drawing on Zeta Alpha's own research into self-optimizing systems for deep research. Aimed at engineers, product and tech leads, and anyone building agentic systems, LLM applications, or AI dev tools, it's a practical, discussion-led evening at the University of Amsterdam's LAB42. Expect to leave with concrete ideas and a room full of people working on the same hard problems.