Speaker
Zhou Yu
Founder @Arklex.ai and CS Professor at Columbia University - Co-Director of the DAPlab
Full speaker profileAbstract
90% of the enterprise agents are stuck in POC. We don't know when the agents are good enough to deploy without a systematic testing and evaluation process. This presentation will share best practices, our practical dev tools, and lessons from the field to help you break out of the demo loop and build AI agents that deliver real business impact. In the talk, we will showcase an example of multi-turn agent simulation tool for automated agents
Main Takeaways:
- Simulation is critical for agent evaluation to reduce manual testing and improve agent quality pre-production and during iterations
- Multi-turn eval is necessary in agent eval to guarantee agent quality
More from the program
Your Newest Employee Has Sudo Access: Blueprints to Stop AI Insider Threats
Adrianna Valle · Klaviyo
Prompt to Prod: Engineering an Autonomous SDLC at Scale
Andrew Swerdlow · Roblox
Finetuning Your Embedding Model for Better Search: Learning from Agentic Search at Dell
Rachel Shalom · Dell Technologies
See all sessions ›