This repository is a portfolio-first MLOps monorepo on Google Cloud, created by completing and then modernizing a full MLOps course curriculum into production-minded projects.
Rather than treating the course as the end goal, I use it as raw material to:
- modernize legacy labs for current GCP defaults
- introduce proper CI/CD separation (PR vs deploy)
- apply Cloud Run and Artifact Registry best practices
- curate selected labs into ML infrastructure–quality portfolio projects
- Google Cloud ML Infrastructure Engineer
- Machine Learning Engineer (Platform / Serving / Pipelines)
- Cloud Run model serving
- CI/CD with Cloud Build (PR validation + branch-based deployment)
- Artifact Registry image management
- Progressive refactoring from course lab → portfolio-quality service
- Local reproducibility on macOS Apple Silicon (M2) using conda + Docker
- End-to-end ML lifecycle awareness (training → serving → orchestration)
Path:
Section3-CloudBuild-CICD/cloudrun-ml-models/coupon-recommendations-v2/
Highlights:
- production-style structure (
src/,tests/,scripts, configuration) - health endpoints for Cloud Run
- Cloud Build build → push → deploy pipeline
- deployable on Google Cloud Run
Path:
Section3-CloudBuild-CICD/cloudrun-ml-models/coupon-recommendations/
Kept intentionally for learning fidelity and to show evolution. Includes separate CI and CD configurations:
cloudbuild.pr.yaml— build and test onlycloudbuild.deploy.yaml— build, push, and deploy
Section3-CloudBuild-CICD/— Cloud Run, Cloud Build, and ML model serving (most portfolio-relevant)Section4-ContinuousTraining-Airflow-Composer/— continuous training and orchestrationSection5-7-VertexAI-Development/— Vertex AI training, batch prediction, explainabilitySection6-Kubeflow-Pipelines/— experiments and pipelinesSection7-Feature-Store/— feature store examplesSection8-GenAI/— supplementary GenAI labsdocs/— setup, workflows, and CI/CD strategyscripts/— reusable infrastructure helpersarchive/— preserved pre-modernization snapshots
dev— course work, refactors, experiments, frequent deploymentsmain— curated, stable, portfolio-ready implementations only
CI/CD is structured to support:
- PR validation (build + test only)
- branch-based deployment for
devandmain
Detailed setup and workflows are documented separately:
- GCP project & IAM setup:
docs/SETUP_GCP.md - Local development (macOS M2):
docs/LOCAL_MAC_M2.md - Progressive lab workflow:
docs/PROGRESSIVE_LABS.md - CI/CD trigger strategy:
docs/TRIGGERS_DEV_MAIN.md
This repository builds on publicly available course material for educational purposes. Refactored implementations reflect independent engineering decisions and modernization work.