Kubeflow
An open source AI platform for Kubernetes that supports self-hosted ML workflows, distributed training and model management under Apache 2.0.
Kubeflow is a self-hosted AI platform for teams that run machine learning workloads on Kubernetes. It brings model development, training and production workflows into a modular stack that can run on a local laptop, on-premises infrastructure or a cloud Kubernetes cluster. It's open source under Apache 2.0.
Platform teams can use individual projects or deploy the full Community Distribution. That flexibility suits organizations building a shared environment for AI practitioners, with tools for both interactive experiments and repeatable jobs.
- Kubeflow Notebooks provides interactive development environments for AI, machine learning and data work.
- Kubeflow Trainer supports distributed model training and LLM fine-tuning with frameworks including PyTorch, MLX, HuggingFace, DeepSpeed and JAX.
- Kubeflow Pipelines builds and deploys portable machine learning workflows on Kubernetes.
- Kubeflow Katib automates hyperparameter tuning, early stopping and neural architecture search.
- Kubeflow Hub tracks models, versions and artifact metadata between experimentation and production.
Kubeflow's focus is the infrastructure around AI workloads: teams can choose components for their own Kubernetes environment and keep the same workload code across local, on-premises and cloud deployments. The Spark Operator also manages Spark applications on Kubernetes, while the Central Dashboard connects authenticated web interfaces across Kubeflow and other ecosystem components.
