University project. A medical pre-diagnosis web application built for the Senegalese context. I trained a RandomForest classifier (scikit-learn) to predict likely conditions from symptoms, and an LLM (Llama 3 via Groq) turns the prediction into a plain-language explanation. Responsive Tailwind frontend, containerized with Docker and deployed on HuggingFace Spaces.
My part: I trained the RandomForest classifier, built the FastAPI REST API exposing the model, integrated the LLM explanation layer, designed the responsive Tailwind frontend, and containerized and deployed the app with Docker.
This application was built as a uni project in a team of 2. It allows for clients to find cleaning personnel, filter by service type (cleaning the house, ironing clothes, etc), schedule, and rating. Both parties verify identity on signup for safety purposes. Deployed on Render
My part: I was in charge of the backend, so I designed the routes, database architecture, data models and auth. I later deployed it through Render
Trained gemma-4-e4b-it-4bit (downloaded from HuggingFace) with LoRA adapters through MLX. Built a synthetic dataset of 57 examples with 5 validation tests. After training, the model went from failing the task to succeeding consistently
My part: Solo project: dataset creation, training pipeline, evaluation
Uni project built in a team of 2. The platform manages a multi-sport club end to end: members, teams, training sessions, attendance, dues and competitions. It enforces real business rules and computes metrics directly in SQL.
My part: I built the Flask backend using blueprints and parameterized SQL queries. I implemented the business rules (minimum age for senior teams, one team per sport, transfer history via entry/exit dates) and SQL-computed metrics like attendance rate.
This application was a uni project made with a team of 3. It's a news platform with role-based access (editor/admin), category filtering and full article management. Everything is persisted in a MySQL database.
My part: I built the API that allows full CRUD of users, articles and categories, secured the routes (only allowing access to certain routes if the user has the right role), connected the MySQL database with PDO, ensured data validation, XSS prevention and proper consumption of the API by the client-side JavaScript.
Uni project built in a team of 4. It allows for full CRUD of books, users, deliveries, borrowing, vehicles, etc. Everything is stored in a MySQL database through DAO objects
My part: I wrote the Livre, Livraison and Annexe classes, connected the database through JDBC, built the DAO classes for Livre and Annexe and designed the Main scenario to test the classes. I also proposed the UML class diagram that we followed to create the different classes and wrote the README.md to describe the project.
This is a personal project I worked on to learn how to code in Rust. It simulates left clicks at a fixed interval and automatically switches slots to eat after some time has passed. It listens on a different thread for a keyboard shortcut to stop running. Everything is configurable through modifying constants.
My part: This is a solo project. I implemented every functionality, made it configurable and wrote a detailed README.md for it on GitHub.
Academic project pairing a React frontend with a Flask API. A Leaflet map displays transit stops and the user's geolocation, with data served from the backend.
My part: I worked across the stack: built the Flask API and its endpoints, set up CORS protection, and developed the React frontend consuming the API along with the Leaflet map integration (stops display and geolocation).
Android notes management app built in a team of 6, with local persistence through Room. Notes are created, edited and stored on-device.
My part: I owned the architecture and data layer (Room entity/DAO/repository), built the shared RecyclerView adapter with double-tap handling, and managed the team's Git integration.
Uni project built in a team of 4. A modular CLI application for managing a school's classes, subjects, students and grades. Data is persisted through CSV file parsing and the project is split into separate modules for each domain
My part: I built the class management module and CSV parsing logic. I also maintained the GitHub repo, tested all features and recorded the demo video