CNCF Graduates Kubeflow for Production AI on Kubernetes

The Cloud Native Computing Foundation has graduated Kubeflow, giving the open source AI and machine learning platform CNCF’s highest maturity designation as enterprises move more AI workloads into production.

Kubeflow runs on Kubernetes and provides infrastructure for a range of AI operations, including data processing, model development, distributed training, fine-tuning, inference and model serving. It allows companies to use the platform across public cloud, private cloud and hybrid environments without tying AI operations to a single vendor.

The graduation milestone matters because enterprise AI has moved far beyond the experimental stage. Companies now need to operate models reliably at scale, creating demand for infrastructure that can support AI development from initial data work through active use cases.

Kubeflow brings those various stages under a common Kubernetes-based framework. The platform can be used by data scientists, AI and ML engineers and infrastructure teams, reducing the need for separate systems as a model moves toward production.

The project’s usage figures demonstrate its impressive reach. Kubeflow’s Python packages have recorded nearly 260 million downloads from the Python Package Index. More than 6,600 contributors from over 1,000 organizations have participated in the project, and its repositories have earned more than 33,000 GitHub stars. NVIDIA, Red Hat, Spotify and Bloomberg are among the companies that have used Kubeflow subprojects.

Kubeflow began at Google in 2017 and entered CNCF as an incubating project in 2023. Since then, it has developed into a larger collection of tools for running AI workloads on Kubernetes.

Graduation Requirements

Graduation requires more than adoption. Kubeflow underwent an independent security audit and created a formal steering committee for project governance. It also adopted CNCF’s Code of Conduct and maintains a Core Infrastructure Initiative Best Practices Badge for secure software development.

The platform also connects with other cloud native technologies, including Prometheus, KServe, Feast, Kueue and Istio.

Kubeflow’s roadmap now extends into some of the most compute-intensive areas of enterprise AI. Planned work includes greater support for LLM orchestration, post-training and fine-tuning, advanced data engineering and AI agents.

Other planned developments show an effort to make AI infrastructure easier to use. The Kubeflow community is working on Kale 2.0, an SDK designed to turn annotated Jupyter notebooks into production pipelines and support Apache Spark. KServe development includes resources for distributed LLM serving through OpenAI-compatible APIs, while Kubeflow Notebooks v2 is being redesigned around a declarative architecture intended to improve security and multi-tenancy.

The key point here for enterprise tech teams is the convergence of AI operations and cloud native infrastructure. Kubernetes became a core platform for managing containerized applications, and now Kubeflow’s graduation demonstrates how that foundation is being extended to the operational requirements of AI. For companies investing in AI, that shift creates another option for building AI infrastructure without committing the full lifecycle to a proprietary platform.