Kubeflow Ecosystem: Past, Present, Futu...
summary
Kubeflow 1.10 is released and ships a model registry UI integrated into the central dashboard, a renamed training operator now Kubeflow Trainer with a v2 Trainer API and LLM blueprints, and Katib LLM hyperparameter tuning so LLMs get the same tuning flows as traditional models. Spark Operator becomes a core component, most components enforce a PSS restricted profile, Pipelines 2.4 closes many v1/v2 gaps (parameterized resource limits, fixed subtask outputs), website gets dark mode and contributor cleanup, and the roadmap targets model storage (OCI/S3), model catalog/cards, Pipelines 2.5 security fixes, Notebooks 2.0 workspaces, Feast integration, SDK unification and Helm installs.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:04Kubeflow history and releases
- 3:37Kubeflow 1.10 release
- 4:44New features in 1.10
- 6:56Security and pipeline improvements
- 8:21Release process updates
- 9:03Website and contributor experience
- 11:57Future 1.11 planning
- 13:21Kubeflow roadmap
related talks
31mKubeflow 2026 State of the Union: Orchestrating the Full ML Lifecycle
32mKubeflow Ecosystem: What’s Next for Cloud Native AI/ML and LLMOps
14mKubeflow: Where ML Meets Kubernetes
21mIntroducing Kubeflow (w. Special Guests Tensorflow and Apache Spark)
32mKubeflow Deep Dive
24mThe Good, the Bad, and the Missing Parts of Kubeflow
36mIntroducing Kubeflow (w. Special Guests Tensorflow and Apache Spark)
09mKubeflow Pipelines v2: The Next Generation of MLOps on Ku...