Back to all tools
Kubeflow
Kubernetes-native ML platform — pipelines, training operators, and model serving at scale.
0Free
MLOps
Overview
Kubeflow is the Kubernetes-native open-source ML platform for building, deploying, and managing ML workflows at scale, with pipelines, notebooks, and model serving.
Key Features
- Kubeflow Pipelines for reproducible ML workflows
- Training Operators for PyTorch, TensorFlow, XGBoost distributed training
- KServe for standardized model serving with auto-scaling
- Katib for hyperparameter tuning on Kubernetes
- JupyterHub integration for notebook environments
- Multi-user isolation with namespaces
Real-World Workflows
End-to-end ML pipeline on Kubernetes
- 1Define pipeline components as Python functions with @component
- 2Compose them into a pipeline with @pipeline
- 3Compile and upload to Kubeflow Pipelines UI
- 4Trigger runs manually or via CI/CD on new data
Getting Started
# Deploy Kubeflow on an existing Kubernetes cluster
kubectl apply -k 'github.com/kubeflow/manifests/example?ref=v1.8.0'
# Install SDK
pip install kfp
# Define a pipeline component
from kfp import dsl
@dsl.component
def train(data: str) -> float:
return 0.95Compare Alternatives
See how Kubeflow stacks up against similar tools.