Skip to content

Contributing to ML/MLOps Portfolio

Thank you for your interest in contributing to this portfolio! This project demonstrates end-to-end MLOps practices across 3 deployed ML services.

Getting Started

Prerequisites

  • Python 3.11+ (3.12 also supported)
  • Docker & Docker Compose
  • Make utility
  • pytest for running tests

Before contributing, please review: - Architecture Documentation: Understand the system design - Deployment Guide: Learn deployment and monitoring procedures - Deployment Evidence: Multi-cloud deployment verification

Setup

  1. Clone the repository 

    git clone https://github.com/DuqueOM/ML-MLOps-Portfolio.git
cd ML-MLOps-Portfolio

  2. Install Dependencies We use make to manage dependencies across all projects: 

    make install

  3. Verify Environment Run the test suite to ensure everything is working: 

    make test

Development Workflow

1. Project Structure

The portfolio consists of three main microservices: - BankChurn-Predictor/ (FastAPI + Scikit-learn) - NLPInsight-Analyzer/ (FastAPI + FinBERT / TF-IDF) - ChicagoTaxi-Demand-Pipeline/ (FastAPI + PySpark + Dask)

2. Dependency Management

We use requirements.in for direct dependencies and pip-compile for locking. To add a package: 1. Edit requirements.in in the specific project folder. 2. Run: 

pip-compile requirements.in
3. Re-install: pip install -r requirements.txt

3. Running Locally

Use the unified Docker stack for testing integration: 

make docker-demo

Code Standards

Python Style

  • Formatter: black
  • Linter: flake8
  • Type Checking: mypy
  • Imports: isort

Run the linting suite before committing: 

make lint

Testing

  • Unit Tests: Required for all new logic (tests/test_*.py).
  • Integration Tests: Required for API endpoints.
  • Use tests/integration/test_demo.py for cross-project validation.
  • Ensure all services pass health checks and prediction tests.
  • Coverage: Must remain above 85% (actual: 90–98%, 395+ tests, Codecov verified).

Run integration tests: 

# Start demo stack
docker compose -f docker-compose.demo.yml up -d

# Run tests
pytest tests/integration/test_demo.py -v

# Tear down
docker compose -f docker-compose.demo.yml down

Commit Messages

We follow Conventional Commits:

  • feat: New feature
  • fix: Bug fix
  • docs: Documentation
  • chore: Maintenance (deps, build)
  • refactor: Code restructuring

Example: 

feat(bankchurn): add probability calibration to StackingClassifier

Pull Request Process

  1. Create a branch: feat/my-new-feature
  2. Commit changes ensuring make test passes.
  3. Open a PR targeting main.
  4. Ensure the CI pipeline (Tests, Lint, Docker Build) is green.
  5. Request review.

Development Process & AI Transparency

This portfolio was developed using AI-assisted tools (Cursor / Cascade) for code generation and boilerplate acceleration. All architectural decisions, project selection, MLOps pipeline design, infrastructure choices, and system integration were made by the author.

What AI tools were used for: - Boilerplate code generation (FastAPI endpoints, test scaffolding, Dockerfile templates) - Documentation drafting and formatting - Code refactoring suggestions and performance optimization patterns

What the author owns and maintains independently: - All architectural and design decisions - MLOps pipeline design (CI/CD workflow, security scanning strategy, coverage enforcement) - Infrastructure choices (Terraform modules, Kubernetes manifests, monitoring stack) - Model selection, experiment design, and hyperparameter tuning strategy - System integration, debugging, and production operations - Docker optimization and multi-stage build design

This transparency reflects the industry-standard practice of leveraging AI tools as productivity accelerators while retaining full engineering ownership.

License

By contributing, you agree that your contributions will be licensed under the MIT License.

Last Updated: March 2026 — v3.5.3