Active build
BriefBridge
Turn chaotic case files into findable records, in seconds instead of hours.
Overview
Insurance-defence firms receive case material as one giant PDF that staples together hundreds of pages of medical reports, court filings, insurer letters, and witness statements in no particular order. BriefBridge ingests that compiled PDF, detects document boundaries, classifies each resulting document, and files it under the case. The web app then gives the firm a case workspace with per-document review, full-text search, and a timeline view.
What it involved
- A Python FastAPI processing brain with five stages: extractor (PyMuPDF plus Tesseract OCR), boundary detector, splitter, a Claude-powered classifier, and an embedder.
- Multi-tenant by organization with audit logging on everything.
- A shared package of literal-union status enums that keep the TypeScript and Python sides from drifting across the service seam.
- A Turborepo monorepo: a Next.js web app on Supabase and Prisma, plus the FastAPI service.
Stack
- Turborepo
- Next.js
- Supabase
- Prisma
- Python FastAPI
- Claude