Yohan Abeysinghe

I am a first year PhD student in Electrical and Computer Engineering at the University of Maryland. My research philosophy is defined by a "proof-over-heuristic" mindset. While modern deep learning thrives on empirical scaling, I am interested in the theoretical foundations of AI—specifically using optimization theory and high-dimensional geometry to explain why models generalize.

Previously, I was a Research Associate at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), working with Dr. Salman Khan and Prof. Yinon Rudich. I received my B.Sc. (First Class Honours) from the University of Moratuwa. I am also a former International Mathematical Olympiad (IMO) and International Physics Olympiad (IPhO) medalist.

I am currently focused on bridging the gap between physics-based constraints and data-driven learning, specifically investigating:

• The theoretical foundations of VLM explainability and interpretability
• Algorithmic frameworks for multi-agent coordination and reasoning
• Stochastic sampling for modeling uncertainty in physical systems
• The mathematical limits of neural operators and high-dimensional optimization

Email  /  CV  /  Google Scholar  /  Github

• Bronze Medal, International Mathematical Olympiad (IMO), Brazil, 2017
• Bronze Medal, Asia Pacific Mathematics Olympiad (APMO), Mexico, 2018
• Honourable Mention, International Physics Olympiad (IPhO), Israel, 2019
• Honourable Mention, Asian Physics Olympiad (APhO), Australia, 2019
• Honourable Mention, International Mathematics Competition for University Students (IMC), Bulgaria, 2022
• World Rank 163, IEEEXtreme 15.0 Programming Competition, 2023
• Honourable Mention, International Mathematical Olympiad (IMO), South Africa (2014), Thailand (2015), Hong Kong (2016)