playanagrams.com ↗
A multiplayer word game built end to end — real-time play, ELO matchmaking, server-side anticheat, and hand-drawn art and sound. Past 5,000 matches.
Brighterway YC S24
Building AI document infrastructure for healthcare. Halved our inference cost while making the document splitter more accurate, took client capacity 10× with HIPAA-compliant ingestion APIs, and shipped a voice agent that takes patient histories alongside physicians.
Multiple sclerosis lesions are rare enough that models starve for data, so I synthesized them — turning 177 real lesions into 10,000+ training samples. The network trained on them beat the published state of the art by 8%. First-author paper in progress.
Yale Center for Biodiversity & Global Change
Built species-distribution reporting for mol.org, including seasonal modelling that lets ecologists trace migration across the year.
VinBigData
On the computer vision team, designing and testing deep learning methods for medical image segmentation — including a 3D U-Net built from scratch to segment liver lesions on CT scans.
A multiplayer word game built end to end — real-time play, ELO matchmaking, server-side anticheat, and hand-drawn art and sound. Past 5,000 matches.
An adaptation of the Minecraft minigame, rebuilt in Unity — you shoot eggs to dissolve the floor out from under everyone else, last player standing wins.
Website and participant portal for Yale's official hackathon — 1,000+ hackers, $10,000+ in prizes each year.
Group scheduling for Yale students and faculty, without the back-and-forth.
Report generation for species distribution models, so ecologists can see how a model performs and how species move.
A method for synthesizing MS rim lesions on QSM scans to fix a brutal class imbalance — and a ResNet trained on them that beats QSMRim-Net.
A 3D U-Net built from scratch for liver lesion segmentation on CT, with the computer vision team.
Yale University
BS Computer Science
BS Statistics & Data Science
May 2025
Teaching assistant — S&DS 265, Introductory Machine Learning
Always up for a conversation about backend systems, applied ML, or whatever you’re building.