I'm an AI engineer and founder in New York. I built and shipped SplitNShare on my own: an AI receipt-splitting iOS app with ~1,000 downloads and a 5.0★ rating, running on a Claude-backed extraction pipeline with its own eval harness.
By day I'm a data engineer at Trajektory, building the TV-viewership and social data pipelines that pro sports teams use to value their sponsorships. I studied Computer Science and Data Science at UChicago.
AI receipt splitting for iOS, built and shipped solo.
~1,000 downloads · 5.0★ · 550+ bills worth $50K+ split
Scan a receipt, assign items to people, and settle up. Textract handles OCR; when a scan fails validation, a Claude fallback with deterministic totals checks repairs it, recovering 76% of failed scans. Serverless AWS backend (Lambda, Cognito, S3) running for about $3/month.
Open-source benchmark for receipt and document extraction.
Compares OCR-only, OCR + LLM repair, and direct vision-model pipelines on field accuracy, line-item F1, latency, and cost per document, with CI-gated evals that block regressions.
Website for the NYC football club I captain.
Squad pages, fixtures, and results for a club in the Bowery Premier League. Match data syncs from the league's Google Sheets, so results update without a deploy.