Agentic RL happy hour at ICML 25 with Veris AI

​Grab a drink, trade ideas, and meet the team behind Veris AI. We’re bringing together researchers who push the boundaries of post-training for agentic systems.

​Why swing by?

• ​Real talk on post-training: Swap successes and pain points with peers who live and breathe reward models, curriculum design, and adversarial testing.

• ​Live demos and lightning chats: See how Veris AI’s high-fidelity simulation platform lets enterprise agents learn by doing, facing realistic user flows, edge cases, and regulatory constraints before they ever hit production.

• ​Fun first: Expect RL-themed cocktails, Vancouver-local bites, and an easygoing vibe perfect for sparking collaboration (or friendly debates over sample-efficiency numbers).

​If you’re attending ICML, refining a new DPO variant, or just excited about making autonomous systems safer and more capable, this is your spot.

​About Veris AI

​Veris AI builds high-fidelity simulation environments where enterprise agents learn by doing. We replace brittle prompt tweaking with realistic user flows, edge cases, and regulatory constraints so models can practice, fail safely, and refine themselves through reinforcement learning. Backed by an $8.5 million seed led by Decibel Partners and Acrew Capital, our team of researchers and engineers is focused on making agentic AI robust for the real world.