A premium monorepo system connecting photographers, hosts, and guests. Drag-and-drop uploads, instant vector matching search, automated watermarking, and adaptive HLS video streaming.
Three specialized user experiences driven by a unified reactive pipeline.
For photographers. Concurrent drag-and-drop batch uploaders to Cloudflare R2, live payment trackers, watermark configuration, and automated guest proofing list CSV exports.
For event organizers. Filter by ceremony folders, star shortlist proofing selections (max 100 limit), and Stripe checkout modals to instantly trigger R2 watermark removal.
For guests. Snap front-facing selfies to locate matches via cosine similarities. Save results to a chronological timeline vault protected by phone/email verification.
Simulate the guest portal search. Snap or choose a selfie and see our face embedding match database files instantly.
Built with enterprise scalability, sub-second latency, and cloud isolation in mind.
Python service hosting the buffalo_l facial vector model on ONNX Runtime, generating 512-dimensional arrays from incoming base64 streams.
PostgreSQL vector database hosting HNSW indexes using vector_cosine_ops operator for similarity retrieval.
Redis-backed daemon workers processing watermark resizing, thumbnail compilation, and automated webhook triggers.
FFmpeg background processing pipelines compiling uploaded raw videos into HLS adaptive bitrate segments (1080p, 720p, 480p).