AI Potluck
Contribute
Back to Gap Map Infrastructure / ML orchestration

Kubeflow Pipelines

Kubeflow
open source / Overall score: 4.2(strong)

Platform for building and running portable machine-learning workflows on Kubernetes. A pipeline is composed in Python from components whose inputs and outputs are typed ML artifacts - Model, Dataset, Metrics, ClassificationMetrics - and each component runs as a container step with an accelerator type and limit set on it. Runs, their artifacts and their lineage are tracked in ML Metadata.

Built on Argo Workflows, which is excluded from this category: the difference is the typed ML artifact system in the KFP SDK, which Argo has no equivalent of. Verified 2026-09-15 via the repository, its recursive tree and the PyPI download API.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(kubeflow/pipelines carries the backend, the frontend and the SDK)
core-gated
ungated(a 5,632-entry recursive tree carries a root LICENSE, every other license file under a third_party/ path, and no path matching enterprise, ee, commercial or proprietary)

Apache-2.0 under CNCF governance with no vendor and no paid edition. The 5,632-entry recursive tree carries one governing LICENSE and every other license file is a vendored third-party dependency.

  • https://github.com/kubeflow/pipelines recorded 2026-09-15

    Repository page for kubeflow/pipelines: Apache-2.0 license, public and unarchived, described as 'Machine Learning Pipelines for Kubeflow'. Establishes the license and that the source is public.

  • https://api.github.com/repos/kubeflow/pipelines/git/trees/master?recursive=1 recorded 2026-09-15

    Full untruncated recursive tree of the default branch, 5,632 entries. Licenses are the root LICENSE; every other license file sits under third_party/ or frontend/third_party/. No path matches enterprise, ee, commercial or proprietary. A tree lists paths; it is cited for that and for nothing about the vendor's offerings.

  • https://www.cncf.io/projects/kubeflow/ recorded 2026-09-16

    Kubeflow's project page on cncf.io, establishing CNCF hosting rather than vendor ownership, so no commercial party is positioned to withhold functionality from the published source.

Adoption

5 high confidence
5.0

The PyPI download API reports 11,622,421 downloads of kfp in the trailing 30 days, above the >10M floor for level 5. kfp is the pipeline SDK and the product's declared distribution channel; the figure is read from the download API rather than a rendered page, so it can be recomputed and challenged.

Capability

3 medium confidence
3.0

Admitted on the ML-job-model limb, and this is the record that shows why the limb is worth having: KFP runs on Argo Workflows, which this category excludes, and the difference is one file - sdk/python/kfp/dsl/types/artifact_types.py defines Model, Dataset, Metrics and ClassificationMetrics as typed artifacts, which Argo has no equivalent of. Band 3: KFP decides the order and the cluster, and the Kubernetes scheduler places the pod.

Verified 2026-09-15