Kubeflow Pipelines v2: The Next Generation of MLOps on Ku...
summary
Kubeflow Pipelines v2 surfaces metadata and metrics as first-class objects with a standardized ontology and queryable artifact lineage, adds richer metric capture and comparisons, and streamlines the SDK with an @component annotation and tools to build containers directly from Python functions. The UI is rebuilt to support control flow (loops, conditionals) and to scale to hundreds of steps, and a platform-independent intermediate representation (IR) lets pipelines authored in KFP run on backends like Vertex Pipelines with little or no change; an initial v2 is already available on GitHub.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 0:10What is a Pipeline
- 1:29Kubeflow Pipelines
- 2:58Metadata Metrics
- 5:38SDK
- 6:26UI
- 7:01Intermediate Representation
- 8:04Next Steps
related talks
29mML Engineering Kubeflow
32mA Deep Dive into Kubeflow Pipelines
35mOrganizing data engineering around Kubeflow
31mKubeflow 2026 State of the Union: Orchestrating the Full ML Lifecycle
14mKubeflow: Where ML Meets Kubernetes
36mDemocratization of ML Pipelines and Bringing ML Workflows to Heterogeneous...
18mKubeflow Ecosystem: Past, Present, Futu...
27mKubeflow for enabling AI Powered Drug Discovery and development...