Master's student in Bioinformatics & AI (BIAM) — FSDM, Université Sidi Mohamed Ben Abdellah Biomedical imaging · statistical genetics · computer vision
I build end-to-end pipelines where machine learning meets biology and medicine — from GWAS summary statistics to MRI volumes to live video.
| Project | What it shows | |
|---|---|---|
| 🧬 | pipeline-pneumonia-cvd-colocalization | GWAS×eQTL colocalization pipeline (FinnGen × GTEx, R + Python). Team project: rigorous null result across 74 gene×tissue tests |
| 🏥 | hospital-finder-fes | Hospital finder for Fès region — interactive map + search with geolocation, Python + OpenStreetMap |
| 👁️ | eye-cataract-detection | HOG+SVM vs fine-tuned YOLOv5 on cataract photographs — full comparison with executed notebooks |
| 🎥 | video-object-tracking-portfolio | 5-module journey: background subtraction → HOG/YOLO → optical flow → DeepSORT → SiamRPN++ |
| 🏷️ | image-classification-tsa | Transfer learning image classification (TFMS) — VGG19, MobileNetV2 fine-tuning, bilingual docs |
Languages: Python · R · SQL · MATLAB ML/DL: PyTorch · scikit-learn · YOLOv5/v8 · coloc / susieR Domains: GWAS & eQTL analysis · radiomics · DICOM/medical imaging · OpenCV video pipelines Practice: reproducible pipelines · honest reporting of negative results · bilingual EN/FR documentation
- Honest science — my colocalization project's main result is a well-evidenced negative; I think that's a feature, not a bug.
- Provenance — every dataset credited and documented from raw source to result.
- Reproducibility — scripts, requirements, and regenerable outputs in every repo.