🎓 Computer Engineering graduate (Université Hassan I, Morocco) 🔬 Research Assistant @ UM6P Analytics Lab — multi-camera perception & deep learning 🌍 From Burkina Faso · based in Rabat, Morocco
I am a Computer Engineering graduate working as a Research Assistant at the UM6P Analytics Lab, where I build and deploy multi-camera video-surveillance systems powered by deep learning — pedestrian detection, multi-object tracking, person re-identification across non-overlapping cameras, and action recognition, all under real-time constraints.
My research focuses on a single question: why do models that excel on benchmarks fail once deployed in the real world, and how can we make them robust? A detector tuned on one camera degrades on the next; re-annotating every new view does not scale. This gap — between academic datasets and real deployment — is what I work on, and what I want to pursue at doctoral level.
I am currently applying for PhD positions in computer vision and machine learning, with a focus on robust perception, representation learning, and label-free adaptation.
- Perception in the wild — detection, tracking, and person re-identification across cameras
- Robustness under distribution shift — domain adaptation, source-free and test-time adaptation
- Self-supervised & foundation models — learning useful representations from unlabelled data
- Deployment-oriented evaluation — knowing when a model can be trusted, and selecting models without labels
- World models & generative video — as tools for building and stress-testing perception
- A. U. Toe, O. Bourja. Person Re-Identification Beyond Benchmarks: A Survey on the Path from Academic Datasets to Real-World Multi-Camera Deployment. In preparation — targeting Artificial Intelligence Review.
- A. U. Toe, O. Bourja. Deployment-Oriented Selection of Low-Resolution Face Recognition Models from Unannotated Videos. In preparation — targeting IEEE AVSS.
- Research Assistant — UM6P Analytics Lab, Rabat (Feb 2026 – present) Multi-camera perception on real video: detection (YOLOv8), tracking, re-identification (OSNet), action recognition (R(2+1)D).
- R&D Intern — Atlantic Smart Technology, Kénitra (2026) Time-series forecasting of financial flows and KPIs from enterprise data.
- Research Intern (AI for Health) — FST, Université Hassan I (2025) Deep learning for early cardiac-event detection, integrated into a clinical-facing app.
Languages: Python · C · C++ · SQL Deep Learning: PyTorch · TensorFlow / Keras · scikit-learn Computer Vision: OpenCV · YOLO · OSNet · object detection, tracking, re-identification Data & Tools: NumPy · Pandas · Git · Linux · GPU clusters Foundations: strong applied mathematics — linear algebra, probability & statistics, optimization
- Applying for PhD positions in robust computer vision & representation learning
- Writing two first-authored papers on the benchmark-to-deployment gap in re-identification
- Deepening my study of self-supervised learning, domain adaptation, and foundation models
⭐ Feel free to explore my repositories — they reflect my journey in machine learning and computer vision.