AI agents are moving fast—from single-turn demos to complex, tool-using systems deployed in production environments. But building agents that actually work in the wild demands more than clever prompts—it requires infrastructure, observability, evaluation, and hard-won lessons from real deployments.

In this panel, Comet CEO Gideon Mendels leads a discussion with engineering and product leaders shaping the next generation of agent systems. With deep experience across tooling, architecture, and applied AI, the panelists will explore what it takes to design, debug, and evaluate agents that operate reliably and safely.

Expect a candid conversation grounded in real-world use cases—from orchestration patterns and human-in-the-loop systems to how teams are using tools that accelerate their daily work.