How the Data Center Shapes the Workload | Nscale | Ray Summit 2026
Everyone talks about the software layer or the chips. Equally important is the control plane deciding which GPU a workload lands on, and when it's trustworthy enough to accept traffic. At Ray Summit 2026, Oscar Savolainen, Senior Staff AI Engineer at Nscale, walks through Nscale's fleet manager as a state machine, from power-on and burn-in to Kubernetes admission: how silent stragglers are filtered out before a customer ever sees the node, and how InfiniBand and NVLink topology shape the units of scale Ray workloads actually run on. You'll leave understanding how the data center itself shapes the workloads you run on it. Liked this video? Check out other Ray Summit breakout session recordings: https://www.youtube.com/playlist?list=PLZNaJyYZRPdo Subscribe to our YouTube channel to stay up-to-date on the future of AI: https://www.youtube.com/c/anyscale š Connect with us: LinkedIn: https://www.linkedin.com/company/joinanyscale/ X: https://x.com/anyscalecompute Website: https://www.anyscale.com/