Kubeflow

What it does

Kubeflow is an open source platform made of modular projects that run data, AI, and machine learning workloads on Kubernetes. Developers and data scientists bring code, datasets, and model configurations, and get back trained models, tuned hyperparameters, managed ML pipelines, and model metadata.

Teams combine the subprojects they need: Trainer for LLM fine-tuning and distributed training, Katib for hyperparameter search, Pipelines for repeatable workflows, Notebooks for interactive development, and Hub for indexing models, versions, and artifact metadata. A central dashboard connects the authenticated web interfaces of these components. The same code can run on a laptop, on-premises, or in any cloud.

Kubeflow suits AI practitioners, platform administrators, and organizations running Kubernetes who want a modular, extensible stack for data and AI work. It is a Cloud Native Computing Foundation project. It requires a Kubernetes environment, so teams without cluster experience should expect to invest in infrastructure setup.

What it can do

• Distributed LLM fine-tuning and AI model training across frameworks such as PyTorch, JAX, and DeepSpeed (Kubeflow Trainer)
• Automated hyperparameter tuning, early stopping, and neural architecture search (Kubeflow Katib)
• Building and running portable, scalable machine learning workflows (Kubeflow Pipelines)
• Running interactive development environments for AI, ML, and data workloads (Kubeflow Notebooks)
• Indexing and managing ML models, versions, and artifact metadata (Kubeflow Hub)
• Specifying and running Spark applications on Kubernetes (Kubeflow Spark Operator)

Works with

• Kubernetes
• PyTorch
• HuggingFace
• JAX
• XGBoost
• DeepSpeed
• Megatron
• MLX

Before you choose

• Requires Kubernetes knowledge or infrastructure; Kubeflow is a platform of components rather than a turnkey hosted service.
• Subprojects are modular, so teams must select and combine tools themselves.
• Costs depend on the compute and infrastructure where it runs; verify cluster and hosting costs for your environment.

Listed under

• Coding
• Sandboxes and infrastructure
• Developers
• Hyperparameter tuning
• Kubernetes
• Machine learning
• MLOps
• ML pipelines

Get started

A Get Started link on kubeflow.org leads to the official installation documentation.

1. Prepare a Kubernetes environment where Kubeflow will run.
2. Choose the subprojects you need, such as Trainer, Katib, Pipelines, or Notebooks.
3. Open the Get Started page on kubeflow.org to follow the official installation documentation.

Verify the compute and infrastructure costs of the Kubernetes environment where you deploy Kubeflow.

We couldn’t find published prices for Kubeflow.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Kubeflow ↗www.kubeflow.org

Company website

kubeflow.org: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.