Projects

Meta-learning framework for few-shot medical image segmentation, enabling U-Net-based models to generalise to new organs or modalities from as few as one labelled example.

Real-time web application that translates hand gestures into text using Mediapipe skeleton tracking and a CNN classifier, built for the Youth For Challenge association.

Automated football match analysis system that detects and tracks players from video, generates tactical maps, and visualises team performance metrics through an interactive dashboard.

Intelligent travel planning chatbot for Morocco, combining RAG and fine-tuned LLMs to deliver personalised itinerary recommendations through a conversational interface.

Machine learning system that helps runners find their race photos automatically using person re-identification and bib-number OCR.