Data Science Template

The data-science template is the core of this meta-project. It follows a modular structure inspired by best industry practices.

Folder Structure

├── .github/          # GitHub Actions CI/CD workflows
├── .gitlab/          # GitLab CI/CD pipelines
├── app/              # Streamlit dashboard
├── api/              # FastAPI model serving
├── data/             # Local data storage (Git ignored)
│   ├── external/
│   ├── interim/
│   ├── processed/
│   └── raw/
├── dockerfiles/      # Optimized Docker images
├── models/           # Trained models (Git ignored)
├── notebooks/        # Jupyter notebooks for EDA and experimentation
├── src/              # Main project source code
│   └── {{ package_name }}/
│       ├── core/     # Core logic and utilities
│       ├── model/    # Model training and prediction
│       └── ...
├── tests/            # Pytest test suite
├── Makefile          # Automation tasks
├── pyproject.toml    # Dependency management
└── README.md         # Project documentation

Core Concepts

  • Separation of Concerns: Logic is in src/, UI is in app/, API is in api/.

  • Reproducible Environment: uv.lock ensures every developer uses the exact same versions.

  • Docker Ready: Multi-stage Dockerfiles for optimized production builds.

The Makefile

The Makefile is your main entry point for automation:

  • make dev-install: Setup virtual env and project dirs.

  • make check: Run Ruff (lint) and Mypy (types).

  • make test: Run all tests with coverage.

  • make api: Launch the FastAPI server.

  • make app: Launch the Streamlit dashboard.