Architecture
Face detection runs in your browser. Search uses public engines without ever sending your image. DMCA notices file under your own name through templates that meet 17 USC §512(c)(3) statutory requirements. Each part of the architecture is designed so your biometric data lives on your side, not ours.
Breakthrough 1
ML runs in your browser
ArcFace face detection runs client-side via ONNX Runtime Web. Your photo never leaves your device. Only a 2KB vector is sent.
Breakthrough 2
Users are the crawlers
The browser extension turns every user into a sensor. Coverage scales with the network — every install adds the pages they already visit, building a distributed index no centralized scanner can match.
Breakthrough 3
DMCA filed under your own name
Notices are sent from the affected person’s email through templates that meet 17 USC §512(c)(3) statutory requirements. Legally stronger than third-party-filed claims, and the platform replies to you directly.
The Stack
Layer
Provider
Notes
Frontend
Cloudflare Pages / Vercel
Edge-cached worldwide
Database + Auth
Supabase
EU + US regions, encrypted at rest
Vector Search
Supabase pgvector
Embedding similarity, no images
File Storage
Cloudflare R2
Evidence packs only — never your face
Face Embedding
Client-side ONNX
Runs in your browser — your photo never leaves
Face Search
Yandex / Bing / Google Lens / FaceCheck
Public reverse-image engines
Email Alerts
Resend
TLS in transit, EU + US regions
DMCA Notices
Counsel-reviewed templates
Filed under your name, not ours
Durable Workflows
Inngest
Idempotent retries, full audit trail
Crawling
Distributed worker mesh
No image upload required