About Me
I’m a junior at Duke University studying CS and Electrical/Computer Engineering. My background is in building data pipelines, full stack web apps, and algorithmic programming (USACO Gold). I'm interested in working on 0→1 consumer-facing projects at high-growth companies. I value ownership over projects and prefer to work on backend systems.
Last summer ('24), I worked on backend systems at Ramp, developing end-to-end flows to automatically evaluate and set credit limits. I also scoped + implemented a core application to support an internal application used by ~200 employees.
The previous summer ('23), I designed and developed a data pipeline at Hotplate, a Series A startup backed by Y Combinator. I worked with APIs and cycled proxy connections for webscraping to analyze Instagram accounts for market intelligence purposes.
Projects
Pattern Tracking with Poi
I built an application that helps performers track the movement of their props with computer vision.
In the past, I've taught how to spin poi, a prop that relies on recreating patterns called "flowers," which can be represented by parametric equations.
Computer vision is used to extract the relative coordinates of the poi and regression is used to evaluate how accurate the patterns that performers create with their poi are to their graphical counterparts.
In the video, the model detects the location of the poi (blue) and uses gradient descent to find the best fit for the parametric equation of the flower (red).
The model is built with YOLOv5 trained on a custom dataset and the computation and analysis is done in Python.
Python, Pytorch, Numpy, Matplotlib
Pattern Tracking with Poi
I built an application that helps performers track the movement of their props with computer vision.
In the past, I've taught how to spin poi, a prop that relies on recreating patterns called "flowers," which can be represented by parametric equations.
Computer vision is used to extract the relative coordinates of the poi and regression is used to evaluate how accurate the patterns that performers create with their poi are to their graphical counterparts.
The model is built with YOLOv5 trained on a custom dataset and the computation and analysis is done in Python.
A video is viewable on the desktop version of this site.
Python, Pytorch, Numpy, Matplotlib
WalletWatchers
WalletWatchers is a financial tracker that aggregates information across multiple bank accounts to generate spending and budgeting insights.
Created API endpoints to fetch financial data the Plaid API, which is stored using SQLite. The website is built with NextJS and TailwindCSS.
Schema was designed with E-R diagrams and testing datasets were developed and generated in Python using Faker to test app functionality.
NextJS, TailwindCSS, SQLite, Plaid API, Python
Quarcade
Quarcade is a fullstack online game platform that supports real-time interactions between users. Players can join lobbies using a generated room code, and can vote to play a multiplayer version of the game Bananagrams.
User data is stored in MongoDB and Websockets are used to facilitate communication between users. The website wireframe was developed in Figma.
This project was developed through Stuyvesant's Google Mentorship program. At the end of our project, we presented our project to Google employees.
ReactJS, MongoDB, Express, Node.js, Socket.io, Figma
Hobbies
Fire Spinning
I started performing with fire poi this past year. The heat of the flames and the sound of the fire engulfing the poi puts me in a state of flow that I can't find anywhere else.
In my free time, I enjoy choreographing routines to music and practicing new tricks. I love the mathematical aspect of spinning poi and the challenge of creating and performing new patterns and sequences.
I've been spinning poi for almost 5 years now, taught over 100 people how to spin, and performed dozens of times.