In full-scale production, the challenge isn’t just model quality—it’s predictable latency, reliability under heavy concurrency, and sustainable unit economics at scale.  In this session, Jacob Yundt (CoreWeave) breaks down a practical readiness framework for running thousands of concurrent agents in real environments. From power and cooling constraints to network contention and tail latency, he’ll unpack what fails first in real-world environments and the infrastructure patterns that help you  keep performance predictable.  You’ll walk away with clear principles to evaluate platforms, pressure-test your architecture, and move from pilot to production with confidence in the agentic era.