Kubeflow runs the ML lifecycle on Kubernetes: Pipelines for orchestrating steps, training operators for distributed jobs, and KServe for scalable model serving - reusing your existing cluster and GitOps.
Why teams use it
- Runs ML workloads on the Kubernetes you already operate
- Reusable, versioned pipelines for training/eval
- Autoscaling model serving with KServe
Install it
# Install Pipelines standalone on an existing cluster
export PIPELINE_VERSION=2.2.0
kubectl apply -k "github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=$PIPELINE_VERSION"A first working config
A minimal pipeline with the KFP SDK.
from kfp import dsl
@dsl.component
def train(lr: float) -> float:
return 0.94
@dsl.pipeline(name='train-pipeline')
def pipe(lr: float = 0.01):
train(lr=lr)Run and verify
kubectl -n kubeflow port-forward svc/ml-pipeline-ui 8080:80
# compile and submit
python -c "from kfp import compiler,dsl; compiler.Compiler().compile(pipe,'pipe.yaml')"Mistakes to avoid early
- Underestimating the operational weight of full Kubeflow
- No GPU scheduling/quotas, causing noisy-neighbor issues
- Pipelines with no caching, re-running expensive steps
Want senior help putting Kubeflow to work - or a team bootcamp on it? CodeWizard runs Israel's largest DevOps community and embeds named senior engineers with a transparent hours log. Book a free consult or explore the Academy.