Fadel Mamar Seydou

The project is documented at: https://github.com/FadelMamar/datalabeling * Leading computer vision algorithms development for aerial wildlife census * Achieved a 84.4% mAP:50 and 80.5% F1-score in single class detection * Reduced data labeling time by 30% through active learning * Addressing large class imbalance through continuous learning strategies

* Implemented a WhatsApp chatbot PoC for improved customer support * Achieved 80% satisfaction score upon delivering a 15-hour training on AI productivity tools to 3 IT public workers * Delivered a 16-hour data analysis workshop to 16 public IT system directors

* Implemented Deep learning models to build digital twins of masonry wallsfrom both real-world and synthetic wave propagation data * Achieved 91.45% F1-score for segmentation of sonic waveforms * Simulated sonic wave propagation through walls using Finite-differencesJules

* Leveraged deep learning segmentation models to enhance eye lesions detection in OCT images, for age-related macular degeneration AMD * Achieved a dice score of 87.5% at b-scan level of the 3D OCT volumes * Identified and addressed critical data annotation inconsistencies among experts