Senior Research: Multimodal Deep Learning Theory

I'm fortunate to be given the opportunity to conduct research with Dr. Zico Kolter, Vaishnavh Nagarajan, Paul Liang, and possibly a few others. We are investigating theoretical explanations for the effects of multimodal deep learning models, such as the benefit of training multimodally for a unimodal task, and the robustness of multimodal models. See here for some blog updates

Dense CNN for Chest Disease Identification

In Hack Auton, the machine learning hackathon, some friends and I trained a Dense Convolutional Neural Net to identify chest diseases. By incorporating additional training features and a skip connection, we achieved better AUROC scores than the state-of-the-art CheXNet model. See our paper

Snowflake

Freshman year of college, I participated in a hackathon with some friends where we created an automatic fridge inventory management system that can recommend recipes based on the inventory. How does it work? See demo

Sight2Sound

Another project completed in another hackathon freshman year. This project can potentially help blind people perceive their surroundings. How does it work? See demo

Neural Networks

In my senior year of high school, I experimented with neural networks and succeded in creating a digit classifier and a digit constructor. See demo

Swarm Robotics

In high school, I also joined a research lab investigating swarm robotics. One of my tasks was writing an ant colony simulation. The purpose was to engineer the swarm to be able to adapt on the fly to a changing food map. We used the Hamiltonian Method of Swarm Design to prove that the task could be satisfied by the swarm. See demo, or see the conference presentation.