MLOps Quick Reference Cheatsheet🔗

🐙 Git🔗

git init                          # initialize repo
git add .                         # stage all changes
git commit -m "message"           # commit
git checkout -b feature/branch    # new branch
git push origin branch-name       # push branch
git pull origin main              # pull latest
git tag v1.0.0 && git push --tags # tag a release

📦 DVC🔗

dvc init                          # init DVC in repo
dvc remote add -d gcs gs://bucket # set remote storage
dvc add data/dataset.csv          # track large file
dvc push                          # push to remote
dvc pull                          # pull from remote
dvc repro                         # reproduce pipeline
dvc dag                           # view pipeline DAG

🐳 Docker🔗

docker build -t image:tag .       # build image
docker run -p 8000:8000 image     # run container
docker run -d --name app image    # run detached
docker ps                         # list running
docker logs app                   # view logs
docker exec -it app bash          # shell into container
docker push gcr.io/proj/img:tag   # push to registry
docker system prune -a            # clean up all

☸️ Kubernetes🔗

kubectl apply -f deployment.yaml  # deploy
kubectl get pods -n production    # list pods
kubectl get svc                   # list services
kubectl logs pod-name             # pod logs
kubectl describe pod pod-name     # pod details
kubectl exec -it pod-name -- bash # shell into pod
kubectl scale deploy/app --replicas=5  # scale
kubectl rollout status deploy/app # check rollout
kubectl rollout undo deploy/app   # rollback
kubectl delete -f deployment.yaml # remove

☁️ GCP / gcloud🔗

gcloud auth login                  # authenticate
gcloud config set project my-proj # set project
gcloud container clusters create ml-cluster \
  --num-nodes=3 --region=us-central1  # create GKE
gcloud container clusters get-credentials ml-cluster \
  --region=us-central1             # connect kubectl
gsutil cp file.csv gs://bucket/    # upload to GCS
gsutil ls gs://bucket/             # list GCS
gcloud builds submit --config cloudbuild.yaml  # cloud build

🔬 MLflow🔗

import mlflow
mlflow.set_experiment("exp-name")
with mlflow.start_run():
    mlflow.log_param("lr", 0.01)
    mlflow.log_metric("accuracy", 0.92)
    mlflow.sklearn.log_model(model, "model")

# Load model
model = mlflow.sklearn.load_model("models:/name/Production")

📊 Kubernetes Manifests🔗

# Minimum deployment
apiVersion: apps/v1
kind: Deployment
metadata:
  name: ml-model
spec:
  replicas: 3
  selector:
    matchLabels:
      app: ml-model
  template:
    metadata:
      labels:
        app: ml-model
    spec:
      containers:
      - name: ml-model
        image: gcr.io/project/ml-model:v1
        ports:
        - containerPort: 8000
        resources:
          requests: {memory: "256Mi", cpu: "250m"}
          limits: {memory: "512Mi", cpu: "500m"}

📌 MLOps Maturity Checklist🔗

Level 0 (Manual):
  □ Model trained in notebook
  □ Manual deployment

Level 1 (Automated Training):
  □ Training pipeline automated
  □ Data versioned (DVC)
  □ Experiments tracked (MLflow)
  □ Model registry in use

Level 2 (CI/CD + CT):
  □ CI pipeline (lint + test on every push)
  □ CD pipeline (auto-build + deploy on merge)
  □ CT (auto-retrain on new data/drift)
  □ Model quality gates
  □ Containerized deployment (Docker)
  □ Kubernetes orchestration
  □ Monitoring + alerting (Prometheus/Grafana)
  □ Drift detection (Evidently)

🔗 Key Tool URLs🔗