A full-stack, production-grade football analytics system that transforms raw StatsBomb event data into deep tactical insights, player ratings, match reports, and real-time intelligence — all powered by machine learning. Implements the VAEP framework (KDD 2019 state-of-the-art) alongside Expected Goals (xG) and Expected Threat (xT) models, orchestrated through a 10-phase automated pipeline.

Most clubs and analysts lack the infrastructure to quantify player actions, detect tactical patterns, or deliver insight in real time. This platform closes that gap with a reproducible ML pipeline feeding a live API and interactive dashboard — built end-to-end from research paper to working product.

A hackathon-born digital health platform focused on breast cancer awareness for Moroccan women — built as a portfolio-ready product combining accessible education, empathetic AI guidance, and practical prevention tools. Live and deployed.

Built at a university hackathon and shipped as a real, live product — not a prototype. Features an AI chat assistant that speaks Moroccan Darija, a simulated SMS reminder flow for self-check habits, and localized risk-awareness content.

A real-time computer vision system for detecting, tracking, and analyzing bird movement in aerial video streams — designed to protect aircraft and human lives through proactive AI surveillance. Bird strikes have caused 262+ deaths and $1.2B in annual damage in the US alone; current prevention remains reactive and human-dependent. This project directly targets that gap.

Combines motion analysis and deep semantic segmentation to detect birds, track them individually with intelligent contour merging, and generate professional targeting overlays with dynamic crosshairs and speed estimation — a genuine attempt at closing a real, unsolved aviation safety gap.

A complete student management web platform for higher-education institutions, exposing two strictly isolated interfaces: a powerful admin back-office (full CRUD, analytics dashboards, grade entry, SQL console) and a self-service student portal where each user sees only their own data. Built as a final-year (PFE) project.

Security here is treated as a first-class architectural concern, not a plugin — every protective layer was implemented manually, without relying on a framework's built-in guardrails. The academic objective: prove security-by-design is achievable from first principles.

The first open-source, end-to-end NLP pipeline for Moroccan Arabic (Darija) — from raw web scraping to preprocessing, model training, and live inference. Darija is spoken by 40M+ people daily yet remains one of the most NLP-neglected dialects in the world, blending Classical Arabic, French, Tamazight, and Spanish influences.

Built specifically to address the absence of NLP tooling for a mixed, code-switching dialect spoken by tens of millions — a genuinely underserved corner of applied NLP.

A distributed flash-sale simulation system evaluating the performance and reliability of Redis-based concurrency control under high request loads — modeling a real-world scenario where thousands of users compete for limited stock simultaneously.

Uses Redis as a central in-memory datastore to guarantee atomicity and consistency, with a full benchmarking suite measuring system behavior as concurrent load scales — a hands-on deep dive into distributed systems fundamentals.

An end-to-end deep learning pipeline for automated COVID-19 detection from chest radiographs — exploring the full ML lifecycle from data preparation to model comparison, explainability, and interactive deployment. Answers a critical clinical question: can a model trained on public chest X-rays reliably distinguish COVID-19 from Normal and Viral Pneumonia, and can we verify it's learning the right features?

Motivated by RT-PCR's ~20% false-negative rate in early-stage COVID-19 (Kucirka et al., 2020). Grounded throughout in published clinical AI literature (Chowdhury et al. 2020; Roberts et al. 2021, Nature Machine Intelligence), with an explicit focus on explainability over raw accuracy — the project's entire thesis being that trust requires transparency.