Alan John Varghese

I’m a PhD candidate at Brown University working on scientific machine learning advised by Prof. George Karniadakis. My research focuses on building ML models that combine deep learning with mathematical structure from physics, numerical methods, and Hamiltonian dynamics. Before Brown, I completed a Dual Degree (integrated Bachelor + Master of Technology) in Aerospace Engineering at the Indian Institute of Technology Madras and worked on modeling of thermoacoustic systems in Prof. R. I. Sujith's lab.

Email  /  Google Scholar  /  Github

Research

I’m interested in scientific machine learning, physics-informed ML, graph neural networks, dynamical systems, and optimal control. Most of my research focuses on developing structure-preserving machine learning methods for modeling, predicting, and controlling complex physical systems. Some papers are highlighted below.

PI-SONet is a structure-preserving neural operator enabling real-time optimal control in high-dimensional systems.

SympGNN is a graph neural network designed to learn the behavior of physical systems while preserving their underlying geometric and energy-based structure.

TransformerG2G uses transformer-based temporal modeling to learn evolving node embeddings in dynamic graphs, enabling improved prediction and analysis of time-varying network structure.

A fractional order model for thermoacoustic instability is developed. Better alignment with experimental data in terms multifractality is demonstrated.