About Me

Hi, I am a Ph.D. candidate in the Dept. of Psychology at Harvard University, advised by Prof. George Alvarez and Prof. Talia Konkle. I am grateful to be supported by the Graduate Fellowship at the Kempner Institute for the Study of Natural and Artificial Vision. Broadly, I study how the human visual system transforms raw sensory input into structured, meaningful representations and how we can uncover similar mechanisms inside large-scale vision models.

My work bridges cognitive science, computational vision, and mechanistic interpretability to understand mid-level vision: how local elements are grouped into proto-objects, how global shape emerges, and how long-range interactions support object perception in both humans and artificial systems.

• On the cognitive and computational side, I study whether self-supervised learning and vision-language alignment provide the inductive biases to learn holistic shape and show human-like sensitivity to configural relations between object parts.

• On the mechanistic interpretability side, we build a decomposition framework that separates positional and semantic information in Vision Transformers via spectral factorization of the query–key interaction, allowing us to isolate the circuits that govern internal information flow.

• On the brain/neural side, I examine the tuning and spatial organization of visual representations, characterizing how mid-level perceptual grouping mechanisms give rise to object-shape perception in humans and vision models.

For a brief overview of my ongoing research, you can read the short feature article linked on this and this page, and learn more about my background through the Harvard Brain Science Initiative page.

Here is our recent work on holistic processing and shape perception in vision models

Link to paper (NeurIPS 2025), project page, and tweet-thread

Check out our work on modeling cortical topographies

Link to paper (Science Advances 2023) and tweet-thread

Here’s some recent work on computational mechanims underlying mid-level object perception

Link to paper (PLOS Computional Biology, 2025) and tweet-thread

Talk on perceptual signatures of contour integration in humans and deep neural networks

Presented at Vision Sciences Society 2022

Talk on a topographic model of the human visual system using self-organizing constraints

Presented at Vision Sciences Society 2021