- Project: 12 points
- In-class follow-up: 8 points
- UV Library Publication → link to PyPI library → 3 points
- Dependency Injection for Inference → link to personal GitHub → 3 points
- MLOps Lab Work → link to personal GitHub → 2 points
The project must:
- Be hosted on a public open-source Git repository
- Include Pull Requests with review/validation between team members
- Contain Unit Tests, ideally following a TDD (Test-Driven Development) approach
- Include a public Kanban board
- Contain a preprocessing phase, such as:
- In the MLOps lab, the preprocessing step is the extraction of images from a PDF
- It can also be another turnkey AI component (e.g., text detection, OCR, etc.)
- Include a mandatory AI training phase:
- The AI can be trained from scratch (requires a large dataset)
- Or fine-tuned from an existing model (e.g., from Hugging Face)
- Use more than 100 data samples (images, audio, videos, etc.) created by you (not downloaded from the internet) and annotated manually
- Be exposed as a real-time API
- Optionally include a web front-end (not mandatory)
| Item | Description | Points |
|---|---|---|
| Teamwork, Clean Code, Pull Requests, linked commits and cards | Proper commit history | 1 |
| Unit-tested preprocessing phase | 2 | |
| AI training phase | 1 | |
| Annotation phase | 2 | |
| Real-time API deployed in an environment | 2 | |
| Automated integration tests | 2 | |
| Final Presentation (5 minutes) | 2 |
- Purpose: What is your project’s functional value? What problem does it solve for end users? What makes it innovative?
- Demo
- Why are you the best?
- You may include technical details about MLOps, cost, and Time to Market.
- What is missing — both technically and functionally.
Choose the format you believe adds the most value to your project presentation!