Sous le capot : les environnements RL chez Zapier avec Will Brown de Prime Intellect

Agentic AI systems are only as good as the environments they learn in. Join Prime Intellect’s Will Brown and Zapier's Robin Salimans for a behind-the-scenes look at how we're using RL environments and verifier-driven methods to evaluate, optimize, and safely scale agent performance in production. You'll get a candid, technical view of how production-grade AI systems are built and optimized—exploring how thoughtful environment design, data selection, and verifiable reward rubrics can push us toward the ICPC gold medal equivalent of automation. This is a rare opportunity to witness how two leading teams are collaborating to enhance the open AI ecosystem and develop the next generation of agentic systems.

In this webinar, you'll learn:

Building RL Environments at Zapier: Discover how Zapier engineers are designing reinforcement learning environments tailored to real-world automation data, creating the foundation for training and optimizing AI agents that handle production workloads.

The Role of Verifiers and GEPA:Get a deep dive into the Verifiers library (created by Will Brown) and explore GEPA (Genetic-Pareto) as a method for continuous, data-driven prompt optimization that enables safe, scalable evaluation loops.

Real-World Insights from Production:Learn the lessons, wins, and pitfalls from applying RL environments to agents doing real automation work, and understand how this Zapier-Prime Intellect partnership is shaping the future of agentic systems.