Mridul Gupta

My research sits at the intersection of Graph-Based Deep Learning and Resource-Constrained Machine Learning. My research has been recognized at Machine Learning venues such as ICLR and KDD. I am particularly interested in making graph ML practical for real-world and scientific applications, with a focus on scalable methods such as graph condensation that compress large datasets without compromising performance.

Currently advised by Dr. Sayan Ranu and Dr. Hariprasad Kodamana at the Yardi School of Artificial Intelligence, IIT Delhi, my doctoral work explores efficient graph representation and distillation frameworks. My recent work, Bonsai, introduces a model-agnostic, linear-time approach to graph distillation, achieving strong performance across multiple benchmarks while significantly improving computational efficiency.

I am broadly interested in understanding the principles behind scalable and deployable graph learning systems, particularly in scientific domains, and aim to contribute to deployable, accurate, and efficient graph learning frameworks for scientific applications, designed to be accessible and researcher-friendly.