I'm a Clinical Data Scientist building a research career at the intersection of AI and medicine. My work spans medical imaging, graph neural networks, and neuroscience — developing models that can support clinicians with more accurate, consistent, and accessible diagnostics.
- 🔬 Presented at MICCAI 2025 — poster presentation on breast cancer segmentation in DCE-MRI
- 🎓 Currently in the BASIRA 2026 Program (Master GNNs for Rising Stars) @ Imperial College London (I-X), under Prof. Islem Rekik
- 🏆 Designed and ran a GNN competition on Parkinson's detection — 18 forks, live leaderboard, auto-scoring
- 💼 Data Scientist at CAAT (Compagnie Algérienne des Assurances) — building AI-driven risk models
- 🌍 Seeking a PhD position in medical AI — oncology, neuroimaging, or clinical informatics
- 📍 Based in Algiers, Algeria
These projects are currently in progress. Repos will be made public as work matures.
| Project | Description | Methods |
|---|---|---|
| 🧠 Pediatric Brain Cancer Segmentation | Tumor segmentation in pediatric brain MRI using transfer learning from adult models | nnU-Net · Transfer Learning · 3D MRI |
| 🧬 Parkinson's Early Detection | Early-stage PD biomarker detection from multi-modal clinical data | GNNs · Acoustic Features · Classification |
| ⚡ GNN for EEG Stroke Detection | Graph-based modeling of EEG signals for early stroke detection | GNNs · EEG · Signal Processing |
Selective Phase-Aware Training of nnU-Net for Robust Breast Cancer Segmentation in Multi-Center DCE-MRI
Poster Presentation · Daejeon, South Korea · September 2025
- Addressed multi-center variability in DCE-MRI breast tumor segmentation
- Demonstrated that quality-selective training (DUKE + NACT) outperformed larger mixed datasets
- Best validation Dice score: 0.72 using 3-phase nnU-Net 3D full-resolution
- Key insight: curating training data by image quality > training on more data
Parkinson's Disease Detection using Graph Neural Networks
Competition Designer & Organizer · BASIRA 2026 Program · 2026
- Designed a full research competition from scratch: dataset, graph construction, evaluation pipeline, auto-scoring
- Built a live leaderboard with GitHub Actions for automated submission scoring
- Framed PD detection as a node classification problem on KNN + subject-connectivity graphs (195 nodes, 22 acoustic features)
- Challenge attracted 18 forks and active community participation
- Covers GNN concepts: GCN, GAT, GraphSAGE, message passing, class imbalance handling
📎 View Challenge Repository · 🏆 Live Leaderboard
| Project | Description | Stack |
|---|---|---|
| 🩺 MICCAI MAMA-MIA Challenge | Breast tumor segmentation in multi-center DCE-MRI with selective phase-aware nnU-Net training | nnU-Net · PyTorch · NIfTI |
| 🧠 PARK-GNN Challenge | Full GNN competition I designed — PD detection from acoustic voice graphs, live leaderboard | PyG · DGL · GitHub Actions |
| 📊 Customer Churn & Recommender System | Predictive churn model (95% accuracy) + GIS spatial analysis across 39 provinces for Djezzy | Scikit-learn · QGIS · Power BI |
| 🛰️ Urban Green Space Mapping | U-Net segmentation on satellite imagery for Frankfurt green space detection | U-Net · QGIS · Rasterio |
| 📖 Arabic Manuscript Detection | Active learning + OCR for historical Arabic handwritten document classification | TensorFlow · OpenCV · NLP |
| 🎓 WQU Applied Data Science Lab | 8 end-to-end data science projects across scientific computing, ML, and deployment | Python · SQL · APIs |
- BASIRA 2026 — Master GNNs for Rising Stars · Imperial College London (I-X) · Online
- MICCAI Winter School 2025 · MBZUAI, Abu Dhabi
- RISE-MICCAI Summer School 2025 · Diffusion Models & Graph Learning · Online
- 🎓 PhD positions in medical AI, clinical data science, or computational medicine
- 🔬 Remote research collaborations in oncology, neuroimaging, or clinical informatics
- 🧑🏫 Mentorship exchanges — I design challenges, teach, and love learning from others
"The goal is not just to build models that perform, but models that help people."