Accepted to CVPR 2019, one of two primary authors. Research conducted in the Stanford Vision and Learning Lab (SVL), under Prof. Silvio Savarese's guidance in the Jackrabbot robotics team. Developed novel architecture combining attention modules on social and physical features to generate trajectories. Pre-print paper can be found here.

Accepted to CVPR 2019, third author. Research conducted in the Stanford Vision and Learning Lab (SVL), under Prof. Silvio Savarese's guidance in the 3D vision team. Focused on a new point cloud decoder that learns arbitrary topologies. Pre-print paper coming soon.

Published in the Journal of Molecular Biology, one of four primary authors. Research conducted in Prof. Rhiju Das's lab in the Stanford School of Medicine. Funded by SIMR. Paper at https://www.ncbi.nlm.nih.gov/pubmed/26902426.

Organized workshop hosted in CVPR 2018 for an algorithms challenge centered around trajectory forecasting. The challenge, TrajNet, can be found at trajnet.stanford.edu. Workshop link here.

In progress. Focuses on developing faster transformers for the task of large-document summarization using two proposed improvements: convolutional encoders, and strided neighborhood attention layers. Conducted as a research project for CS224N (Deep Natural Language Processing) with two other student researchers. Paper can be found here.

Develops a generative model for predicting poverty from daytime satellite images of a region. Focuses on learning representations of satellite images and transferring these representations to other domains. Conducted as a research project for CS236 (Deep Generative Modeling) with two other student researchers. Paper can be found here.

Improves the accuracy of error classification techniques during genome sequencing using features learned from a recurrent embedding. Conducted as a research project for CS230 (Deep Learning) with two other student researchers. Paper can be found here.

Develops a novel architecture to reconstruct a 3D MRI across multiple slices using a single slice as input. Learns translations between consecutive image slices. Conducted as a research project for CS231N (Convolutional Neural Networks) with two other student researchers. Paper can be found here.

Attempts to improve a baseline word embedding by utilizing recurrence in groups of words. Conducted as a research project for CS224U (Natural Language Understanding) with two other student researchers. Paper can be found here.

Develops a gambling agent for the Stanley Cup Playoffs by training a game winner classifier using historical game data and modeling a multi-state Markov Decision Process. Conducted as a research project for CS229 (Machine Learning) and CS221 (Artificial Intelligence) with two other student researchers. Paper can be found here.

Introduces a novel algorithm, combining machine learning and crowdsourcing, to design RNA riboswitches that are stable in silico and in vitro. Research project selected as 1 of 300 nationally for the Intel Science Talent Search. Paper can be found here.
Machine LearningBiology

Develops a new crowdsourcing interface, Eterna3D, to allow for the design of atomically precise 3D RNA folds. Crowdsourced molecules were experimentally synthesized and shown to achieve the required design targets. Paper can be found here.