Developed C++ client for optimized reading of the internal-to-Facebook fork of the Apache Hadoop ORC file format. Benchmarks showed as much as 50% improvement in CPU time on local file reading speed against existing Java implementation.
Providing instruction and assistance to students in introductory computer science courses. I assisted the most students per shift of all lab TAs between January and May 2017.
Developed Recollect, a photo compilation app for automatically organizing photos by date and location into easy-to-share digital scrapbooks.
Developed interfaces to integrate the AMBER molecular dynamics simulation software project with other quantum mechanics simulation packages. My code remains in production in the official AmberTools distribution.
HackMIT winner (Baidu's $10,000 prize for Best Paddle Hack). We use Baidu's Paddle Deep Learning framework to uncover and call attention to online harassment and its biases. I developed our Paddle-powered deep learning model for identifying comments as containing personal attacks. I also wrote backend integration code for making inferences with this model against Facebook comments and Tweets.
Platform for Princeton University students and faculty to buy and sell goods. Developed with React/Redux on the frontend, and Go on the backend. I was the primary frontend developer.
Developed a webapp and award-winning HackPrinceton 2015 project built on Meteor and React that constructs a Markov model from a given user's Twitter timeline to generate similarly-written tweets. I was responsible chiefly for the backend code for retrieving Tweets, generating the model, and generating clean-looking output.
I developed a novel ant colony optimization algorithm for solving the traveling salesman problem that used multiple distinct "colonies" to avoid local maxima. I published a paper demonstrating my algorithm’s significant performance advantages.
React, Redux, Node, Android
My team won the Google-sponsored coding/puzzle/trivia/word-association competition.