Charmed Kubeflow

What it does

Charmed Kubeflow is Canonical's supported distribution of Kubeflow, an MLOps platform for building, training and deploying machine learning and AI workflows. It takes in data science workloads such as training jobs, experiments and inference pipelines, and runs them on Kubernetes across public clouds, private infrastructure, VMs, bare metal or air-gapped environments.

People use it to run parallel experiments, separate training and inference environments, and give data science teams secure multi-user workspaces with role-based access. Its distinguishing points in the source are the Juju-based deployment for consistent operations across environments, and Canonical-maintained support with security updates and upgrades, alongside a modular stack that can include Charmed Kubernetes, Ceph and observability tools.

Charmed Kubeflow suits data science and platform teams running Kubernetes-based machine-learning infrastructure. The software is open source, while Canonical offers commercial support subscriptions priced per node per year. Deployment requires a suitable Kubernetes environment and operational expertise. Confirm supported versions, node counts, and the support scope before selecting a subscription.

What it can do

• Deploy Kubeflow on public clouds, private infrastructure or air-gapped environments
• Run parallel distributed training jobs with Kubernetes-native auto-scaling
• Provide secure multi-user workspaces with role-based access
• Built-in GPU-ready configurations
• Deploy with Juju for consistent, declarative operations across environments
• Support for AKS, EKS, GKE and hybrid or multi-cloud setups

Works with

• Charmed Kubernetes
• Ceph
• Observability tools
• Data lakes
• Inference engines

Before you choose

• Pricing is subscription based per node per year, so cost depends on node count; no amounts are shown in the provider’s documentation.
• Deployment and operations expect Kubernetes and Juju experience.
• The free open source distribution includes upgrades and security updates, but enterprise support is a paid subscription.

Listed under

• Coding
• DevOps and observability

Get started

The page directs readers to get in touch; a Kubeflow installation tutorial is also linked.

1. Set up a Kubernetes environment such as MicroK8s, Ubuntu Kubernetes or a managed cloud service.
2. Install Juju and follow the Kubeflow installation tutorial.
3. Configure multi-user workspaces and role-based access for data science teams.

Open-source software can be deployed independently. Commercial support subscriptions are priced per node per year; confirm the support scope and compute costs separately.

Source: canonical.com/mlops/kubeflow, checked 30 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 30 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Kubeflow | MLOps | Canonical ↗canonical.com

Company website

canonical.com: product, pricing and policy pages

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 30 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.