How KubeRay Is Evolving for Massive AI Workloads  | Ray Summit 2025 @anyscale
How KubeRay Is Evolving for Massive AI Workloads  | Ray Summit 2025  @anyscale
Uploaded November 2025 | Updated November 2026, 21 hours ago
At Ray Summit 2025, Aaron Liang from Google and Jui-An Huang from Anyscale share the latest advancements in KubeRay aimed at dramatically improving the user experience for teams running Ray on Kubernetes.

They introduce a series of major RayJob enhancements designed to simplify and harden the end-to-end workload lifecycle. These include:

Deletion policies for predictable and controllable cleanup

Cron scheduling to support recurring and automated Ray workloads

Sidecar mode for easier integration with surrounding services and tooling

Background status checks that improve reliability and reduce operational overhead

Together, these upgrades make Ray job orchestration on Kubernetes more intuitive, more manageable, and more robust for both developers and operators.

Attendees will gain practical insight into how these new features streamline Ray workloads in production—and how KubeRay is evolving to make large-scale distributed computing easier than ever.

Liked this video? Check out other Ray Summit breakout session recordings youtube.com/playlist?list=PLzTswPQNepXllnU0C36WtkC0dqkAoDulh

Subscribe to our YouTube channel to stay up-to-date on the future of AI! youtube.com/c/anyscale

đź”— Connect with us:
LinkedIn: linkedin.com/company/joinanyscale
X: https://x.com/anyscalecompute
Website: anyscale.com
How KubeRay Is Evolving for Massive AI Workloads  | Ray Summit 2025Production-Ready Distributed Inference with Ray Serve | Anyscale | Ray Summit 2026Petabyte-Scale Video Curation on CoreWeave | CoreWeave | Ray Summit 2026Ray Direct Transport: RDMA Support in Ray Core | Ray Summit 2025Maximizing Compute Efficiency on Anyscale | Ray Summit 2025The PARK Stack: The LAMP Stack of the AI Era | Ben LoricaProduction-ready Distributed Inference with Ray ServeAgentic Workload Inference at Scale: ByteDance’s AIBrix & DeerFlow | Ray Summit 2025Real-Time Search & Recommendation Systems for AI Commerce | Anyscale | Ray Summit 2026Exabyte-scale Streaming Iceberg IO with Ray, Flink, and DeltaCAT | Ray Summit 2025Marin: Open Development of Open Foundation Models | Ray Summit 2025End-to-End GenAI Orchestration with KubeRay | Ray Summit 2025
Anyscale |

How KubeRay Is Evolving for Massive AI Workloads | Ray Summit 2025

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER