/RunThrough/
Enterprise race photography platform that uses AWS AI to match faces and race numbers — so runners stop scrolling and start downloading.
RunThrough had thousands of race photos per event and runners with no fast way to find themselves. We developed an end-to-end platform on Next.js 14 and Supabase where a runner uploads one selfie — AWS Rekognition finds every photo of them across the gallery, ranked by confidence. AWS Textract reads race bibs. Results in seconds, not hours.
/Project Goals/
Kill the manual photo-hunting experience and replace it with AI search that works fast enough that runners actually use it.
The brief was clear — runners were not buying photos because finding their own was too painful. The platform needed face recognition, race number OCR, watermarked downloads, a real-time admin dashboard, and an architecture that could handle thousands of images per event without breaking. Every feature had to serve that one goal: get the runner to their photo fast.
/Results/
Full AI photo discovery live in production — face search, bib OCR, watermarked downloads, and real-time indexing across every event.
Delivered a complete platform covering public gallery, AI search, and admin control panel. New photo uploads trigger automatic AI indexing via S3 webhook. Admins see live stats on downloads, search queries, and indexing progress through Socket.IO and Supabase Realtime. Role-based access, database-level Row Level Security, and JWT auth cover the security layer end to end.
Next.js 16AWS RekognitionAWS TextractSupabaseTypeScriptSharpAWS S3Vercel