kubeflow/kubeflow

Machine Learning Toolkit for Kubernetes

What it solves

Kubeflow provides a foundation of tools to build AI platforms on Kubernetes, simplifying the management of the entire AI lifecycle. It addresses the complexity of deploying, scaling, and managing machine learning workflows across different environments.

How it works

It operates as a composable and modular system of Kubernetes-native projects. Users can either deploy the entire Kubeflow Community Distribution or use individual subprojects independently to meet their specific needs.

Who it’s for

It is designed for AI practitioners, platform administrators, and development teams who need scalable and extensible tools to support their AI use cases.

Highlights

  • Composable, modular, and portable architecture
  • Backed by an ecosystem of Kubernetes-native projects
  • Supports every stage of the AI lifecycle
  • Flexible deployment options via subprojects or a full community distribution

Related

  • Project
    kubeflow/trainer

    A Kubernetes-native distributed AI platform for scalable LLM fine-tuning and training across various frameworks like PyTorch, JAX, and XGBoost.

  • Project
    kubeflow/pipelines

    A platform for building and deploying reusable, scalable end-to-end machine learning workflows on Kubernetes.

  • Project
    kubeflow/katib

    A Kubernetes-native AutoML project that provides automated hyperparameter tuning, neural architecture search, and early stopping for any ML framework.

  • Project
    kubeflow/community-distribution

    A community-driven distribution of the Kubeflow Platform that provides manifests for installing a complete ML lifecycle management system on various Kubernetes clusters.

  • Project
    labring/sealos

    An AI-native Cloud Operating System built on Kubernetes that unifies development, deployment, and scaling for AI applications and SaaS platforms.