Comprehensive Guide to AI/ML Project Life-Cycle Using Git and GitHub¶

Overview¶

Effective use of Git and GitHub is essential for collaboration and maintaining a robust codebase in AI/ML projects. This guide outlines best practices for managing the life-cycle of your project using Git and GitHub, ensuring reproducibility, scalability, and seamless collaboration.

Steps in the Life-Cycle¶

1. Open a JIRA Issue¶

• Open a JIRA issue to track your work and get the issue number. If necessary, create sub-tasks and use their issue numbers for more granular tracking.

2. Create a Branch¶

• Create a branch following the naming standard described in the Git Branch Naming Standards. Including the JIRA number in the branch name aids in navigation across the repository.

3. Start Exploratory Analysis¶

• Begin your exploratory analysis using a Jupyter notebook. This allows for quick prototyping and visualization. Before sharing your notebook, use the Black formatter and the Ruff linter to ensure code quality. Black ensures consistent code formatting, and Ruff helps with sorting imports, including docstrings, and other linting tasks.

4. Evaluate Exploration Outcomes¶

• Unsuccessful Experiment: If your exploration/experiment does not yield a viable solution, document the process and findings in the notebook. Share the notebook, keep the branch open but unmerged, and open a Pull Request (PR) for review. Close the JIRA issue.
• Successful Experiment: If your experiment is successful and needs to be tested in production, proceed with the next steps to transition from exploratory code to production-ready code.

5. Transition to Python Script¶

• Move your successful exploratory code from notebooks to Python scripts (.py). Design your classes, methods, and modules carefully. Ensure proper documentation, formatting, linting, testing, commenting, refactoring, and error handling to prepare the code for production.

6. Document Your Code¶

• If the code is moving to production, document your class and methods using MkDocs. Refer to Build Your Python Project Documentation With MkDocs for guidance. Incorporate the generated documentation into the docs folder.

7. Open a Pull Request¶

• Open a Pull Request, undergo peer review, merge the branch into the production branch, close the JIRA issue, and delete the branch.

Life-Cycle Flow Chart¶

---
title: AI/ML Project Life-Cycle
---
flowchart TB
    A(Open a JIRA Issue) --> B(Create a Branch)
    B --> C(Start Exploratory Analysis)
    C --> D{Is the Experiment Successful?}
    D --> |No| E(Document and Share Notebook)
    E --> I(Open a Pull Request)
    D --> |Yes| F(Transition to Python Script)
    F --> G(Document Your Code)
    G --> H(Open a Pull Request)
    H --> J(Merge Code into Production)