MLOps-GCP

IaC for MLOps on Google Cloud Platform

Requirements

• A Google Cloud Platform account
• Terraform >= 1.14

Architecture

• Artifact Registry: Repository for MLflow Docker container
• Cloud SQL: MLflow back-end store (PostgreSQL) for experiment tracking
• Cloud Storage: bucket for MLflow artifacts store
• Cloud Run: Serves MLflow container
• Secrets Manager: Securely stores database credentials and sensitive URIs
• IAM: Manages service account permissions to architecture components

Download and Installation

Download

Open Cloud Shell on GCP and run git clone https://github.com/keiferc/mlops-gcp.git.

Installation

# Set variables
cd mlops-gcp/terraform/
cp terraform.tfvars.example terraform.tfvars # fill placeholders

# Enable APIs
gcloud services enable cloudresourcemanager.googleapis.com --project=<PROJECT_ID>
terraform init
terraform validate
terraform plan # check plan logic
terraform apply -target=google_project_service.apis # enable APIs

# Deploy MLflow container to Artifact Registry
terraform apply -target=google_artifact_registry_repository.mlflow
cd ../docker/
gcloud auth configure-docker <REGION>-docker.pkg.dev
gcloud builds submit --tag <REGION>-docker.pkg.dev/<PROJECT_ID>/mlflow-repo/mlflow:latest --project <PROJECT_ID> .

# Deploy infrastructure
cd ../terraform/
terraform plan # check plan logic
terraform apply

Usage

Inspect resources with terraform show and destroy resources with terraform destroy.

View MLFlow UI by running the following on Cloud Shell:

gcloud run services proxy mlflow-server --region=<REGION> --project=<PROJECT_ID> --port=8080

then clicking the Web Preview icon.

Contributing

Requirements

• uv >= 0.10

Installation

uv sync --dev
uv run pre-commit autoupdate
uv run pre-commit install

Guidelines

• Write self-documenting code
• Manage dependencies with uv (e.g., uv add polars)
• Submit pull requests to dev