ESC
Saif Ur Rahman

Saif Ur Rahman

Ph.D. Candidate in Computer Science

University of Illinois Urbana-Champaign

Champaign, IL

I am looking for a research internship in Summer 2027.

News

Aug 2026
PU-HNO submitted to NeurIPS 2026; preprint available on arXiv.
April 2026
WiNeRF accepted at EWSN 2026. [Paper]
Feb 2025
Finalist, Qualcomm Innovation Fellowship North America.
Nov 2024
Passed PhD qualifying exam and Advanced to Ph.D. candidacy.
See all news

Biography

I am a Ph.D. candidate in Computer Science at the University of Illinois Urbana-Champaign, advised by Prof. Elahé Soltanaghai in the iSens Lab.

I received my B.S. in Electrical Engineering with a minor in Computer Science from the Lahore University of Management Sciences (LUMS) in 2022, graduating top of my class. As an undergraduate I worked with Prof. Muhammad Tahir and Prof. Momin Uppal on RF sensing with software-defined radios and passive RFID. After graduating, I spent a semester at LUMS's Centre for Urban Informatics, Technology & Policy, building a transformer-based analysis pipeline for a World Bank–funded urban policy project.

My Research

My research lies at the intersection of wireless systems and machine learning. More broadly, I am interested in using learning to model and understand physical systems, particularly in problems where the structure and constraints of the underlying system need to shape how the machine learning method is designed.

Within this broader space, my Ph.D. research focuses on building physics-grounded machine learning models for wireless propagation. Accurate wireless simulation is difficult because real-world measurements are often sparse, while high-fidelity physical simulation can be computationally expensive and require detailed knowledge of the environment.

I explore how learned representations can bridge this gap by incorporating structure from wireless physics, measurement processes, and simulation into the model itself. My work spans neural fields, neural operators, differentiable simulation, and wireless digital twins, with the broader goal of making high-fidelity wireless modeling faster, more scalable, and useful for real-world decision making.

A recurring theme in my work is learning from imperfect but useful sources of information: sparse real-world measurements, lower-fidelity simulation, or simplified representations of physical environments that would otherwise be too expensive to model explicitly.