Kubeflow Pipelines Operator

Manage ML pipelines as Kubernetes resources using GitOps and declarative configuration.

Get Started See the Code

The KFP Operator provides a Kubernetes-native API for Kubeflow Pipelines. Deploy and manage ML pipelines using kubectl, Helm, and GitOps workflows with Custom Resource Definitions.

Kubernetes-Native

Manage ML pipelines as Kubernetes resources using kubectl, Helm, and GitOps workflows.

Event-Driven

Trigger pipeline runs automatically based on schedules, events, or data changes.

Production-Ready

Enterprise security, observability, and integration with existing Kubernetes infrastructure.

How It Works

Define Pipelines as Code

Create Kubernetes manifests for your ML pipelines and version control them with your code.


apiVersion: pipelines.kubeflow.org/v1beta1
kind: Pipeline
metadata:
  name: training-pipeline
spec:
  image: my-org/ml-pipeline:v1.2.0
  env:
  - name: MODEL_VERSION
    value: "v2.1"

Deploy with kubectl

Use standard Kubernetes tools to deploy and manage your ML workflows.


# Deploy your pipeline
kubectl apply -f pipeline.yaml

# Trigger a run
kubectl apply -f runconfiguration.yaml

# Monitor status
kubectl get mlr,mlrc

Get Started

Installation

helm repo add kfp-operator https://sky-uk.github.io/kfp-operator/
helm install kfp-operator kfp-operator/kfp-operator

Open Source

100% open source and welcomes contributions. Built by Sky’s ML Platform team and used in production.

View on GitHub →

Community

Get help and connect with other users in GitHub Discussions.

Join Discussions →

Issues & Feedback

Report bugs and request features on GitHub Issues.

Report Issues →