The Platonic Universe: Do Foundation Models See the Same Sky?
Mechanistic Interpretability Workshop at ICML 2026 · arXiv:2509.19453
My research is in applied AI: using machine learning to solve problems in the physical sciences, and using scientific data as a testbed for understanding how intelligent systems represent and reason.
Most of my published work applies deep learning to high-energy astrophysics: classifying blazars and X-ray binaries, and characterising the optical variability of active galactic nuclei. My recent work has moved toward foundation models and mechanistic interpretability, asking whether models trained on different instruments and modalities converge on the same representation of the sky, and what that tells us about the models themselves.
I hold an M.Sc. in Physics with Astrophysics from IUCAA and SPPU, and a B.Tech. in Electrical Engineering from Tezpur University. My work spans astrophysics, machine learning, uncertainty quantification, computer vision, and scientific computing.
Selected work
ZTF Monitoring of Gamma-Ray Emitting Narrow Line Seyfert 1 Galaxies
arXiv:2602.22410
Gamma-ray blazar classification using machine learning with advanced weight initialization and self-supervised learning techniques
Monthly Notices of the Royal Astronomical Society 528 (1), 976–986
When I am not working on my research projects I am busy thinking about getting students into research through Psintica
Feel free to reach out to me for collaboration or any queries related to my research work.