I am a PhD student at MIT EECS, where I am advised by Aleksander Mądry. Previously, I spent two great years at Google Research and Microsoft Research, where I worked with Praneeth Netrapalli and Prateek Jain. Before that, I studied CS and Stats at UIUC.

I am broadly interested in developing tools to understand and steer model behavior. In particular, my recent work focuses on understanding how training data and learning algorithms jointly shape model behavior at test time.


Papers

ContextCite: Attributing Model Generation to Context
Decomposing and Editing Predictions by Modeling Model Computation
ModelDiff: A Framework for Comparing Learning Algorithms
Do Input Gradients Highlight Discriminative Features?
The Pitfalls of Simplicity Bias in Neural Networks
Growing Attributed Networks through Local Processes
Number of Connected Components in a Graph: Estimation via Counting Patterns