NYU Tandon School of Engineering
Ali Hamza
PhD Student researching self-supervised learning for autonomous vehicle planning. Focused on how SSL-pretrained vision backbones generalize across different cities and driving conditions within the NAVSIM framework.
Advised at NYU Tandon School of Engineering
Cross-City SSL Generalization Study
Investigating how self-supervised learning backbones (I-JEPA, DINOv2, MAE) generalize across Boston, Las Vegas, Pittsburgh, and Singapore. A three-phase experimental study examining single-city training, leave-one-city-out evaluation, and data efficiency.
View research →I-JEPA + MLP Planning
Initial investigation showing that I-JEPA pre-trained features with a lightweight MLP planning head achieve 82.06% PDMS using only camera inputs, matching multi-modal systems that use both camera and LiDAR with full datasets.
View archive →