about/

Hi, I'm Usha! I'm an incoming first-year PhD student at Harvard, coadvised by Prof. Hima Lakkaraju and Prof. Hanspeter Pfister. I go by she/her and they/them pronouns.

Get to know me better!

- email: usha[_]bhalla[@]g[.]harvard[.]edu

image of me

research/

My research interests generally lie in the realm of machine learning/computer vision interpretability.

* Brown Honors Thesis

Completed an Honors Thesis with Professor Chen Sun, where I studied the interpretability of CLIP's representations of primitive concepts.

[T. Yun, U. Bhalla, E. Pavlick, C. Sun. “Do Vision-Language Pretrained Models Learn Primitive Concepts?” arXiv.]

* Microsoft Research Intern

Developed a 4D-spatiotemporal neural network robust to domain shifts and limited data to identify which scenes of a movie people will recall later on using fMRI scans taken during watch time.