Trace It, Prove It, Scale It: AI Agents in Production

An afternoon in Paris on what it actually takes to run agents in production: traces you can read, audit trails you can defend, and compute economics that survive scale. About the event Join the CoreWeave team for an afternoon on scaling agents in production: real users, real data, and a real invoice! Most startup and scale-up teams in Paris are past the prototype stage and into the hard part: a multi-agent workflow that fails silently, a regulator or a client asking how a model reached a decision, and an inference bill that grows faster than usage. Those are three symptoms of the same gap. You can't prove what you can't trace, and you can't scale what you can't prove. If you own the technology decision, that gap arrives as three conversations you have to win. The engineering one, where a silent agent failure becomes your outage and your on-call rotation. The security and compliance one, where "we will document it later" stalls a French enterprise deal for a quarter. And the finance one, where you explain to your CFO or your board why cost per inference is moving the wrong way while revenue per user isn't. So, we are running three working sessions on closing that gap: live tooling you can reproduce in your own stack, real cases with real numbers, and straight answers on portability, data residency, and TCO. We wrap up the afternoon with a drinks reception enabling networking among the attendees and continuation of the conversation. Apply for your place today. Space is limited, as this is an intimate event. Who this is for: Founders, CTOs, and AI leads at startups and scale-ups running agents or models in production Platform and infrastructure engineers responsible for cost, reliability, and scale ML and AI engineers working on evaluation, observability, and post-training Teams facing audit, procurement, or regulatory scrutiny on how their models behave Join us and leave with: A working picture of how to instrument, trace, and evaluate a multi-agent system A view of what an audit-ready AI pipeline looks like before anyone asks for one Concrete numbers to bring into your own build-versus-migrate conversation