This page catalogs some of the projects that I have worked on in the past, as well as a few that I am working on right now. If you have any questions about these or other projects that I have been involved with, don’t hesitate to contact me.
I am currently working with Adrian Sampson at Cornell on a language called Opal. The language has built-in abstractions to make writing machine learning applications easier.
The research that I am currently doing involves creating typed abstractions for various machine learning workflows. I worked for a while on a type system for features; Opal now has supports type-safe features that mitigate mistakes when working with large data-sets. More recently, I have been working on typed interactions with natural language systems. We want to allow programmers using NLU systems to generate configurations from the types in their program.
Last Second Beach
Last Second Beach, LLC is a small start-up that I had the pleasure of co-founding in 2016. I was brought on by Khalid Ladha and Zacharia Demuth as CTO to provide a technical perspective for the budding travel-technology company. We hoped to create a mobile distribution channel for cruises and one-price vacations, a market segment that has traditionally been slow to adapt to changes in technology.
While the company has since become mostly inactive, we did win first place (and $25,000) at the Cornell School of Hotel Administration at their Business Plan competition.
For our high school senior project, my partner and I designed and created a glove that interprets letters in the American Sign Language alphabet and prints the letters to a computer screen. The translation happens in real time, as a person forms the letters with his or her hand. The glove also allowed the user to play Rock, Paper, Scissors against the computer.
We detected the user’s hand position using custom-made optical bend sensors, which we made from aquarium tubing and photo-diodes. With these sensors, we were able to get a general idea of how much each of the user’s fingers was bending. Hall effect sensors and magnets mounted between the fingers allowed us to get more precise measurements. We used a TI micro-controller to interpret the signals and pass them on to a computer, at which point a Python script did the final translation and printed out the results.
The n-body problem in calculus is often used to show the limitations of symbolic integration. It can be thought of as modeling several planets in gravitational orbit. When there are two planets, symbolic integration suffices to determine their paths; when there are three or more planets symbolic integration becomes impossible.
In order to simulate the motion of these planets, each “step” of each planet’s motion must be calculated as a combination of the gravitational effects of all planets around it. My project partner and I used the Runge-Kutta method of numerical integration to find the paths of any number of planets in orbit. We developed a graphical user interface using vPython and wxPython that allows a user to set initial conditions and view a simulation.
Check out the project here.
Mafia Web App
Mafia is a strategic party game played by many at Cornell. The game is both logical and psychological—sometimes players need to use the facts at their disposal to draw conclusions, and other times they need to “read” other players to determine who is lying.
As co-presidents of the Cornell Mafia club, a friend and I decided to create a web application to make playing the game easier. The each game is run by a “moderator”, but sometimes that person’s job can be fairly difficult. We created a node.js app using Meteor that provided tools to make moderating easier.