CS at UVA (May 2027), machine learning engineer intern at Expedia Group this summer. I work on recommender systems and the data pipelines behind them, mostly for things I actually want to use.
right now
- building Marathon Prep Bot: an LLM writes my training plans and is not allowed to compute anything. A TypeScript pipeline decodes Garmin FIT binaries, CRC-validates them, and owns every number, with 243 tests across 32 suites that fail the build when the model and the pipeline disagree. Richmond, November 2026.
- I wrote about how it works: my marathon coach is a chat window
things I've made
- Shortlist: a two-tower retriever over 89,585 movies behind a Go API. 84.1% HitRate@10 against 73.8% for a popularity baseline, measured on 7,060 cold users held out by a global time cutoff and scored exactly once (code)
- Signal: Spotify revoked the endpoints this was originally built on, so I rebuilt discovery on artist collaboration graphs using what the API still supports. Every pick explains its real connection (code)
- Today: the SwiftUI training app I use every morning. Weight logging, strength tracking with a live muscle map, and coach plans, synced with AES-256-GCM where the decryption key never leaves the phone (code)
- rohansingh04.com: the site everything lives on, a project in its own right (code)
find me: rohansingh04.com · linkedin · rohan.singh04@outlook.com