I’m interested in multimodal learning problems at the intersection of computation and cognition.
I completed my undergrad at Harvard SEAS in 2018, where I concentrated in Computer Science and Mind, Brain & Behavior. I wrote my undergraduate thesis and two publications on Visual Question Answering under Alexander Rush and Stuart Shieber and in close collaboration with Yonatan Belinkov. I’ve completed two research-focused internships at Google Brain (now Google AI) on Google’s Tensorflow and OCR teams. Previously, I was an RA in the Fedorenko Lab at MIT, where I worked on computational models of lexical semantics.
I’m currently a research engineer at Reverie Labs, where I’m working to apply techniques from computational chemistry, natural language processing, and computer vision to develop next-generation brain cancer drugs. I’m also an active contributor to the open science COVID Moonshot effort with colleagues from Reverie. Much of my work employs graph-based deep learning (neural message passing, graph convolutions, etc.) and large-scale language models (e.g., Transformers) for molecular property prediction. I’ve also worked on molecular generation, molecular docking, virtual screening design, and interfaces for human-machine interaction. I’ve written about some of these projects in the Reverie Blog.
I like to read broadly and keep up with the literature. One of the areas I’m most excited about is methods that combine neural architectures with symbolic reasoning. Some of my current areas of interest are:
- Neuro-symbolic learning models
- Probabilistic programming
- Program synthesis
- ML interpretability and bias
- Model generalization and out-of-distribution (OOD) testing
- Few-shot learning
- Bayesian ML