Eslam Abdelaleem
Postdoctoral Fellow. Georgia Institute of Technology.
Salam! I’m Eslam.
I am a physicist exploring the intersection of neuroscience, statistical physics, and machine learning.
My goal is simple: I try to find the simplicity (e.g., equations, patterns, etc.) hiding inside the complexity (e.g., brains!).
During my PhD at Emory University with Dr. Ilya Nemenman, I focused on a fundamental problem in modern science: we have too much data and not enough understanding.
I worked on Dimensionality Reduction, developing principled and information-theoretic, yet applicable using modern machine learning tools, frameworks to extract meaningful representations from large, multimodal datasets. Most of my work focused on the “Information Bottleneck Principle”—finding the sweet spot where we filter out the irrelevant -noise- but keep the variables that are actually relevant -signal- to the system.
Now, as a Postdoctoral Fellow at Georgia Tech working with Dr. Audrey Sederberg, I am putting those theoretical tools to work on real neural circuits. My current work provides tools to bridge the gap between abstract machine learning models and the biological reality of how the brain processes information.
Aside from research, I don’t just research physics; I love to teach it. I believe that if you can’t explain it simply, you don’t understand it well enough.
Currently, I am a Lecturer for Physics of Cognition at Georgia Tech, a course that perfectly blends my two professional loves. There is nothing quite like the “aha!” moment when a student connects a complex equation to a real-world concept, which I try to include in every single lecture, assignment, or project.
Feel free to explore my CV, check out my Code, or take a look at my not-professional-at-all-but-i-love-it Photography.
news
| Dec 10, 2025 | ”‘Periodic Table’ for AI Methods Aims to Drive Innovation.“ Emory University News, December 2025. Feature on JMLR 2025 paper (Deep Variational Multivariate Information Bottleneck). |
|---|---|
| Jul 30, 2025 | “AI Reveals Unexpected New Physics in Dusty Plasma.” Emory University News, July 2025. Feature on PNAS 2025 paper. Covered by 21+ news outlets, podcasts, and blogs. |
selected publications
- JMLRDeep variational multivariate information bottleneck-a framework for variational lossesJournal of Machine Learning Research, 2025
- TMLRSimultaneous dimensionality reduction: A data efficient approach for multimodal representations learningTransactions on Machine Learning Research, 2024
- ArXivAccurate Estimation of Mutual Information in High Dimensional DataarXiv preprint arXiv:2506.00330, 2025