Week 5 Review: CI/CD for ML Projects

Automate ML Model Deployment to Kubernetes with GitOps, ArgoCD, and Helm.

By Sandip Das · · 7 min read

Week 5 Review: CI/CD for ML Projects
💡
Hey — It's Sandip Das 👋

Five days, one pipeline. Today we stitch GitHub Actions, DVC, CML, Docker and GitOps into a single flow for iris-mlops, audit it for the gaps that bite in production, and see where it lands on the MLOps maturity model.

📚 Key Learnings

  • Why pushing kubectl apply from a CI runner stops scaling, and what pulling fixes
  • The four GitOps principles, and the one that actually does the work: continuous reconciliation
  • Argo CD installed on minikube, with an Application pointed at your Day 26 Helm chart
  • A CI job that writes the new image digest into a config repo, so deploys happen without anyone typing a command
  • Where GitOps gets awkward for ML, and what you put in Git when the artifact is a 400 MB model

🧠 Learn

Read the full article on learnxops.com.

Topics: 100 Days of MLOps, MLOps, ML, GitOps, ArgoCD