Kubeflow is the cloud-native AI platform: a set of modular, open-source projects that form a Kubernetes-native stack for data and AI workloads across the whole machine learning lifecycle. Its mission is to bridge the data, AI and cloud-native ecosystems, giving practitioners, platform administrators and decision-makers well-lit paths from experimentation to production at any scale. Components cover notebooks, pipelines, distributed training, hyperparameter tuning and model management, all running on Kubernetes. Kubeflow has nearly 16,000 stars on GitHub and the community project was actively developed in 2026.
Rather than a single monolithic product, Kubeflow is a toolkit of community subprojects you can adopt together or separately: managed notebooks for experimentation, Kubeflow Pipelines for orchestrating ML workflows, training operators for distributed PyTorch and TensorFlow jobs, Katib for hyperparameter tuning and neural architecture search, and a model registry for governing what ships. Everything is Kubernetes-native, so it runs wherever Kubernetes runs, from a local cluster to any major cloud, with the same APIs and scaling behavior. Distributions from cloud providers and vendors package Kubeflow with support for enterprises that want a supported platform.
Kubeflow is free and open source under the Apache-2.0 license, developed by a broad community. You pay only for the Kubernetes infrastructure you run it on; vendors also offer supported distributions.
Kubeflow is for ML platform teams and organizations standardizing on Kubernetes who need an open, portable stack for training, tuning and serving models without locking into a single cloud provider’s AI platform.