Youssef Khomany

– Led end-to-end computer vision and machine learning pipelines from data preprocessing and semi-automated annotation to model training and integration. – Designed, trained, and optimized deep learning segmentation models (U-Net variants, ViT attention-based architectures), achieving F1-scores above 95%. – Developed robust classical vision pipelines combining pupil/iris detection, GMM/K-Means clustering, adaptive thresholding, and morphological post-processing. – Extracted ocular redness biomarkers from scleral vein detection and implemented ML-based dry eye severity classification aligned with clinical standards (EFRON, CCLRU). – Integrated trained models into production software and collaborated within an agile R&D environment alongside software engineers and clinical experts.

– Developed image segmentation pipelines for wave-breaking detection using classical methods (Chan-Vese) and deep learning models. – Trained and evaluated a U-Net convolutional neural network for pixel-wise segmentation tasks. – Applied unsupervised clustering (DBSCAN) to group wave-breaking regions and improve spatial consistency.