Mohamed Abdelfattah

I am a computer vision Ph.D. student at EPFL VITA lab, fortunately advised by Alexandre Alahi.

My research focuses on Representation Learning in Multimodal LLMs and Self-Supervised Transformer Foundation Models to address a central challenge in video understanding: learning temporally grounded, cross-modal representations without relying on costly supervision or brittle alignment objectives. I design predictive pretraining strategies that enable models to reason over video, audio, motion, and language in a unified and transferable way.

I was an AI Research Scientist Intern at Meta Reality Labs (Summer 2025) where I worked with Edoardo Remelli and Shugao Ma on building Multimodal LLMs for smart wearable systems. Previously, I was a deep learning visiting research student at Boston University, collaborating with Sarah Adel Bargal and a computer vision research intern the KAUST Vision-Cair group, working with Mohamed Elhoseiny.

I received my B.Sc. in Computer Engineering from The American University in Cairo in 2022 (highest honors, Dean's list).

  Email  /    CV  /    LinkedIn  /  Google Scholar  /    Github


I am actively looking for internship positions in computer vision, LLMs, and deep learning. Feel free to reach out to me!

profile photo

Publications

OSKAR: Omnimodal Self-supervised Knowledge Abstraction and Representation
Mohamed Abdelfattah*, Kaouther Messoud*, Alexandre Alahi
NeurIPS 2025
Project page
Paper

In one sentence: OSKAR is a self-supervised multimodal foundation model that learns in the latent space by predicting masked multimodal features.

S-JEPA: Joint Embedding Predictive Architecture for Self-Supervised Skeletal Action Recognition
Mohamed Abdelfattah, Alexandre Alahi
ECCV 2024
Project page
Paper

In one sentence: S-JEPA is an instatiation of the Joint Embedding Predictive Architecture (JEPA) for self-supervised skeletal action recognition.

MaskCLR: Attention-Guided Contrastive Learning for Robust Action Representation Learning
Mohamed Abdelfattah, Mariam Hassan, Alexandre Alahi
CVPR 2024
Project page
Paper

In one sentence: MaskCLR improves the robustness of transformer-based action recognition methods against noisy and incomplete skeletons.

Toward contactless human thermal monitoring: A framework for Machine Learning-based human thermo-physiology modeling augmented with computer vision
Mohamad Rida, Mohamed Abdelfattah, Alexandre Alahi, Dolaana Khovalyg
Building and Environment, 2023
Paper

In one sentence: A machine-learning human thermo-physiology model (ML-HTPM) developed to predict thermal response.

Non-intrusive physiological parameters sensing for personalized human thermal comfort prediction
Mohamad Rida, Mohamed Abdelfattah, Alexandre Alahi, Dolaana Khovalyg
18th Healthy Buildings Europe Conference, 2023
Paper

In one sentence: In this study, we applied multi-modal non-intrusive computer vision algorithms to extract personal features such as the clothing ensemble, activity level, posture, sex, age, and skin temperature as a human’s thermal comfort defining parameters.

ArtELingo: A Million Emotion Annotations of WikiArt with Emphasis on Diversity over Language and Culture
Youssef Mohamed, Mohamed Abdelfattah, Shyma Alhuwaider, Feifan Li, Xiangliang Zhang, Kenneth Ward Church, Mohamed Elhoseiny
EMNLP 2022
Project Page
Code
Dataset
arXiv

In one sentence: This paper introduces ArtELingo, a new benchmark and dataset, designed to encourage work on diversity across languages and cultures.

ZeroWaste Dataset: Towards Deformable Object Segmentation in Cluttered Scenes
Dina Bashkirova, Mohamed Abdelfattah, Ziliang Zhu, James Akl, Fadi Alladkani, Ping Hu, Vitaly Ablavsky, Berk Calli, Sarah Adel Bargal, Kate Saenko
CVPR 2022
Project Page
Code
Dataset
arXiv

In one sentence: we take a step towards computer-aided waste detection and present the first in-the-wild industrial-grade waste detection and segmentation dataset, ZeroWaste.

Teaching

Awards


Template adapted from Jon Barron's and Siwei Zhang's websites.