I build projects around machine learning, computer vision, and systems programming. My current focus is turning learning projects into cleaner, reproducible, and better-tested software.
| Project | Stack | What it shows | Status |
|---|---|---|---|
| LightGBM-Cybersecurity | Python, scikit-learn, LightGBM, pytest | Packaged ML pipeline with preprocessing, baseline model, validation reports, runtime artifacts, and tests. | Most production-like project |
| ML-Price-Guesser | C++17, CMake, OpenMP | Gradient-boosted decision tree pipeline for real-estate price prediction, built from scratch with temporal splitting and feature engineering. | Active refactor / testing needed |
| LLM-From-Abs-Scratch | C++17, CMake | Neural-network layers and numerical checks built from scratch, with gradient checks for activation layers. | Early research / learning project |
| Octobin | Python, TensorFlow/Keras, OpenCV, Flask | Real-time waste classification prototype using MobileNetV2 transfer learning and webcam-based correction workflow. | Prototype |
| Infinite-Runner | Python, Pygame | 2D game loop, collisions, assets, scoring, pause/game-over states, and gameplay systems. | Learning / game-dev project |
| Get_What_U_Need | Python, Pygame | Larger Pygame project with state management, asset loading, progression, audio, and fullscreen support. | Collaborative / learning project |
These repositories are forks used for exploration. I do not present them as fully original framework rewrites unless the specific changes are documented in the repository.
These repositories are mainly coursework, exercises, or personal practice. They are kept public to show progression, but they are not the best representation of my current engineering standards.
- LAFF-University-of-Texas
- TP1-Genericite
- TP_ProgC
- TP-Classes-internes
- Paradigme-Programmation
- Exercise---gym
- Garage-LE-PNEU-ORIENTE-OBJET
- temporaire
- AlphaZero-but-better-
- Add reproducible setup instructions to every serious project.
- Keep generated files, IDE folders, datasets, and model artifacts out of Git when possible.
- Add tests for core logic, especially C++ ML code and data preprocessing.
- Document which repositories are original projects, forks, prototypes, or coursework.
- Prefer clear configuration files over hardcoded local paths.
- Add CI where the project can be built and tested reliably.
Languages: Python, C++, Java, C
ML/Data: scikit-learn, LightGBM, TensorFlow/Keras, pandas, NumPy
Systems/Tools: CMake, OpenMP, Git, pytest
Creative: Pygame, OpenCV
Thanks for visiting. The strongest repositories to review first are LightGBM-Cybersecurity, ML-Price-Guesser, and LLM-From-Abs-Scratch.