Hi! I’m a ML Research Fellow working in Computer Vision,  Representation Learning, and SciML. I’ve had the chance to work across leading Institutes (IISER Bhopal) and national labs (DYSL - AI Lab) in India, presently applying my learning in the industry settings. I have few publication at venues including IEEE IJCB (Model Compression and Biometrics), ACM COMPASS (Applied ML), and American Physical Society (Computational Physics).

My work, accepted to IEEE IJCB 2025, aims to compress face recognition (FR) models without losing identity geometry, as real world FR system demands privacy (offline) and should be real time. We introduced two complementary knowledge distillation objectives: instance-level embedding alignment and geometric pairwise relation preservation. The unified method allows richer knowledge transfer between the student and teacher model.

Also a part of my work is applying ML in real world uses cases. Imagine a world where technology not only assists but actively ensures our safety on the roads. At IISER Bhopal, we integrated ML with a proposed dynamic adaptive time window algorithm for driving behavior analysis. The work was done with Prof. Haroon Lone and Dr. Manoj Gulati at SIRL Lab and published at ACM Compass 2024.  

Earlier, my work in computational plasma physics with Prof. Bhaskar Chaudhury at DA-IICT, Gandhinagar, shaped my interest in Scientific Machine Learning. I worked on ML surrogate modeling for plasma simulations, which made me think more carefully about how scientific ML models should behave beyond test accuracy. This means building models that respect the structure of the physical system, remain stable in simulation workflows, and generalize to new settings. 

I am looking for PhD position my research interest includes but is not limited to AI for Science, Explainable AI, and Computer Vision.