Kubeflow has grown into a comprehensive ecosystem that powers end-to-end machine learning on Kubernetes.

In this session, we provide a holistic ""State of the Union"" update across all major working groups. We will walk the audience through the lifecycle of a modern model, demonstrating how the unified SDK and Notebook Workspaces simplify authoring.

We will then trace how these workloads transition into scalable Kubeflow Pipelines, undergo distributed training and hyperparameter tuning via Trainer v2 and Katib, and are finally cataloged in the Model Registry before being deployed via KServe.

We will conclude by outlining the upcoming project roadmap and providing updates on how we can improve core Kubeflow components or build integration bridges to adjacent OSS AI projects.