Introduction
Kubeflow is a Kubernetes-native open-source AI platform suite that helps teams build, deploy and manage scalable machine learning workloads. It offers modular components for training, serving, experimentation and orchestration that can be deployed independently or as a complete reference platform.
Key features
- Modular components (Pipelines, Katib, Notebooks, Model Registry) for composable deployments.
- Containerized task execution, resource isolation and support for multiple training and inference backends.
- Visual dashboard, experiment tracking and data lineage for debugging and auditing.
- Active community and enterprise adoption with established installation and deployment guides.
Use cases
- Building end-to-end ML platforms and managed services on Kubernetes clusters.
- Platform teams that need multi-tenancy, shared resources and model lifecycle management.
- Automating model delivery and monitoring with CI/CD and GitOps workflows.
Technical details
- Built on Kubernetes with support for Helm and manifest-based installation options.
- Language-agnostic workflow definitions via containerized tasks and multi-language SDKs.
- Rich ecosystem of subprojects (Pipelines, Katib, KServe, etc.) for extensibility and integrations.