Kubeflow and Kubernetes are a common pairing in a modern DevOps stack. Here is how they fit together and a practical way to wire them up.
Where Kubeflow fits
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.
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)Where Kubernetes fits
Kubernetes is a container orchestrator: you declare the desired state of your workloads in YAML and controllers continuously reconcile the cluster toward it, handling scheduling, scaling, healing and networking.
apiVersion: apps/v1
kind: Deployment
metadata: { name: web }
spec:
replicas: 3
selector: { matchLabels: { app: web } }
template:
metadata: { labels: { app: web } }
spec:
containers:
- name: web
image: ghcr.io/acme/web:1.0
ports: [{ containerPort: 3000 }]
resources:
requests: { cpu: 100m, memory: 128Mi }
limits: { cpu: 500m, memory: 256Mi }
readinessProbe:
httpGet: { path: /healthz, port: 3000 }
---
apiVersion: v1
kind: Service
metadata: { name: web }
spec:
selector: { app: web }
ports: [{ port: 80, targetPort: 3000 }]Making them work together
- Start with Pipelines + KServe rather than the whole platform
- Use GPU node pools with quotas and taints
- Set requests/limits and use the HorizontalPodAutoscaler for elastic workloads
- Separate config (ConfigMap) from secrets and mount them, don't bake them in
Want this pairing set up properly, with docs and handover? That is exactly the kind of work CodeWizard's embedded senior engineers do - and both Kubeflow and Kubernetes have active community groups you can join for free.