"
# post = "Open Source. Battle-Tested. Community Built.
-
- What is Kubeflow?
+
+ An Ecosystem of Tools AND a Cohesive Platform.
-
- Kubeflow is the foundation of tools for AI Platforms on Kubernetes.
-
- AI platform teams can build on top of Kubeflow by using each project independently or deploying the
- entire AI reference platform to meet their specific needs. The Kubeflow AI reference platform is
- composable, modular, portable, and scalable, backed by an ecosystem of Kubernetes-native
- projects for each stage of
- the AI lifecycle.
-
- Deploy Kubeflow anywhere you run Kubernetes.
-
+ Born from a mission to bridge two distinct worlds—cloud-native infrastructure ("Kube") and machine learning + workflows ("Flow")—Kubeflow was created to solve the "Platform Problem." As AI demands scale, modern ML efforts + are often spent on wrestling with a disconnected portfolio of tools that create friction instead of value. +
+ ++ This unifying mission is alive today in Kubeflow's ecosystem of maintainers and expert contributors, working + closely together on a shared vision of interoperability across projects like Pipelines, Notebooks, Trainer, and + more. With Kubeflow, teams can deploy ML workloads on any Kubernetes cluster, tailoring the environment to meet + their exact machine learning needs. By uniting the Kubernetes and ML communities, the result is clear: Data + Scientists get the centralized experience they need, and Platform Engineers can confidently rely on Kubeflow to + abstract their infrastructure complexity away. + + In the end, the Kubeflow community works tirelessly to reduce operational friction so we all can focus on what + matters: our AI applications. +
+Scroll to learn more, or dive straight into our documentation.
+ Explore the Kubeflow Docs ++ What Workloads Run On Kubeflow? +
++ The Kubeflow ecosystem builds specialized components that teams across the community rely on to tackle + everything from traditional predictive models to complex generative AI: +
+-
+
- Develop (Notebooks): Spin up interactive, collaborative environments to + write code and experiment. +
- Train (Kubeflow Trainer): Fine-tune LLMs and train distributed foundation + models at scale. +
- Optimize (Katib): Automate hyperparameter tuning and optimize advanced + architectures like agentic RAG. +
- Process (Spark Operator): Prepare and process massive datasets natively on + Kubernetes. +
- Track (Dashboard): Monitor complex experiments and visualize runs through a + centralized UI. +
- Serve (KServe): Deploy, monitor, and scale generative and predictive models + directly to production. +
- Manage (Model Registry): Index, version, and manage ML models and artifact + metadata to bridge the gap between experimentation and production. +
- Orchestrate (Pipelines): Unify the entire ML lifecycle, enabling teams to + build, share, and iterate on reusable components. +
+ Kubeflow has plenty of use cases! You can learn more about our GenAI use cases, as well as check out + how our community uses Kubeflow on the Kubeflow blog! +
+ ++ Have a great Kubeflow use case? Join our community calls or open a GitHub PR to get your team's story featured on the official website. +
++ Explore the individual project tiles below to learn more about each tool, or keep scrolling to launch + your first deployment. +
+ +
+
+ + Kubeflow Spark + Operator + aims to make specifying and running Spark applications as easy and idiomatic as running other workloads on + Kubernetes. +
+
+
+ + Kubeflow Notebooks provide + interactive development environments for AI, ML, and Data workloads on Kubernetes. +
+
+
+ + Kubeflow Trainer is a + Kubernetes-native project for LLM fine-tuning and enabling scalable, distributed training across a wide + range + of AI frameworks, including PyTorch, HuggingFace, DeepSpeed, MLX, JAX, XGBoost, and others. +
+
+
+ + Kubeflow Katib is a + Kubernetes-native project for automated machine learning (AutoML) with support for hyperparameter tuning, + early stopping, and neural architecture search. +
++
+
+ + KServe is a standardized + distributed generative and predictive AI inference platform for scalable, multi-framework deployment on + Kubernetes. +
+
+
+ + Kubeflow Model + Registry + is a cloud-native component that provides a single pane of glass for ML model developers to index and manage + models, versions, and ML artifact metadata. It fills a gap between model experimentation and production + activities. +
+
+
+ + Kubeflow Pipelines (KFP) + is + a platform for building then deploying portable and scalable machine learning workflows using Kubernetes. +
+
+
+ + Kubeflow Central + Dashboard is our hub which connects the authenticated web interfaces of Kubeflow and other ecosystem + components. +
++
+ +
+ All unified by the Kubeflow + SDK—a unified set of Pythonic APIs designed to accelerate + time-to-value + without requiring deep Kubernetes expertise. +
+Experience the ease of the Kubeflow SDK
+
+ pip install -U kubeflow
+
+ + Getting Started with Kubeflow +
++ Building real AI products requires tools that adapt to unique workflows. Ready to get more models into + production? Our Getting Started guide outlines deployment paths tailored perfectly to various architectural + needs. Adopters can build a highly customized, production-ready MLOps foundation using robust community + manifests, or leverage a packaged, conformant distribution for a more opinionated deployment. Kubeflow is + commited to giving Platform Engineers the infrastructure control they require while ensuring they meet the needs + of their data science and machine learning counterparts. +
+++ + Read the Getting Started Guide + + ++ "We believe Kubeflow is the gold standard for AI/ML on Kubernetes. We are committed to ensuring it empowers + the next decade of AI and ML workloads." +
+
Trusted by
@@ -98,224 +356,60 @@
- Kubeflow Projects -
-
-
- - Kubeflow Spark Operator aims to make specifying and running Spark applications as easy and idiomatic as running other workloads on Kubernetes. -
-
-
- - Kubeflow Notebooks runs interactive development environments for AI, ML, and Data workloads on Kubernetes. -
-
-
- - Kubeflow Trainer is a Kubernetes-native project for LLMs fine-tuning and enabling scalable, distributed training across - a wide range of AI frameworks, including PyTorch, HuggingFace, DeepSpeed, MLX, JAX, XGBoost, and others. -
-
-
- - Kubeflow Katib is a Kubernetes-native project for automated machine learning (AutoML) with support for hyperparameter tuning, early stopping and neural architecture search. -
-+ Want to see your logo here? Open an issue or reach out on the adopters repo, + open a PR, or create an issue on the website GitHub. +
-
-
- - KServe is a standardized distributed generative and predictive AI inference platform for scalable, multi-framework deployment on Kubernetes. -
-
-
- - Kubeflow Model Registry is a cloud-native component that provides a single pane of glass for ML model developers to index and manage models, versions, and ML artifacts metadata. - It fills a gap between model experimentation and production activities. -
-
-
- - Kubeflow Pipelines (KFP) is a platform for building then deploying portable and scalable machine learning workflows using Kubernetes. -
-
-
- - Kubeflow Central Dashboard is our hub which connects the authenticated web interfaces of Kubeflow and other ecosystem components. -
-+ Join our Community +
++ We are an open and welcoming community of software developers, data scientists, and organizations! +
++ Check out the weekly + community calls, get involved in discussions on the mailing list or + chat with others on the Slack Workspace! +
-
- -
- Join our Community -
-- We are an open and welcoming community of software developers, data scientists, and organizations! - Check out the weekly community calls, get involved in discussions on the mailing list or chat with others on the Slack Workspace! -
- 
+
+