Hi, I'm Harsh, a 4th year undergraduate studying Computer Science and Math at MIT interested in machine learning and computer systems.

I’m currently working at MIT’s Institute for Medical Engineering and Science on interpretable models for clinical time-series data. My research uses ideas from variational inference, knowledge distillation, and switching state-space models to learn simple, interpretable models that still capture meaningful temporal structure.

I have previously worked as a software engineering intern at Jane Street, an intern on Microsoft's M365 Copilot Team. I also cofounded and served as CTO for Y Combinator-backed Martin and coauthored papers in Nature Genetics and NeurIPS.

When I’m not building systems, you can usually find me playing badminton, getting overly competitive over a chess board, or speed cubing to clear my head.