Mahmoud “Mido” Assran is a scientist and member of technical staff on AMI’s founding team. He previously spent nearly nine years at Meta, including as a research scientist with Fundamental AI Research (FAIR).

His research focuses on self-supervised learning and world models, particularly Joint-Embedding Predictive Architectures (JEPA). Rather than predicting every pixel, JEPA learns abstract representations by predicting in latent space. Assran’s work includes I-JEPA and V-JEPA; he is also listed as a speaker at TED AI Vienna 2026.

Assran earned a PhD in electrical and computer engineering at McGill University and Mila, the Quebec AI Institute, with support from a Vanier Scholarship. He remains active in research and education as a NeurIPS area chair, a co-organizer of the ECCV World Modeling Workshop, and a mentor and teacher through initiatives including Meta’s AI Mentorship program and the AI Learning Alliance.