The finale: assemble the whole series — FastAPI, Docker, GitHub Actions, ECR, Terraform, EKS, MLflow, Prometheus/Grafana, and drift-triggered retraining — into one platform, and trace a change from git push to a promoted model serving traffic. Part 11 of the MLOps series.
Stop building infrastructure by hand. Provision the VPC, EKS cluster, ECR, S3, and the GitHub OIDC role with Terraform — reproducible, reviewable, and easy to destroy on demand. Part 8 of the MLOps series.
Take the inference container to a real cluster: Deployments, Services, Ingress, ConfigMaps, Secrets, and an HPA — with the readiness probe we've been building since part two finally doing its job. Part 7 of the MLOps series.
Automate the gate: test the model API, build the image, scan it for CVEs with Trivy, and push to Amazon ECR — with OIDC instead of long-lived keys and the git SHA as an immutable image tag. Part 4 of the MLOps series.
Design and deploy a production-grade 3-tier web app on AWS with Terraform, ECS Fargate, RDS Multi-AZ, and a GitHub Actions CI/CD pipeline. The complete build, the reasoning, and the repo you can run yourself.