Ingests 2,000+ grants, scores them for institutional fit, and drafts proposals in the VC’s own voice.
Next.js 16
React 19
Drizzle
Neon
Better Auth
Sonnet
Vercel
//FIG.12 — stratify
A production grant-intelligence engine: thousands of grants ingested and scored for fit, then voice-matched proposal drafts generated automatically — delivered and demoed to the Vice Chancellor.
//Role
Architect & engineer
//Status
Live / Production
//Access
Live / public
//Problem
Grant discovery is a manual, time-starved process. The institution needed thousands of opportunities triaged for fit and a head-start on proposals that actually sound like the people who write them.
//Architecture
A daily ingestion pipeline pulls 2,000+ grants and corporate programs into a Neon/Drizzle store; a Sonnet scoring pass stratifies them by institutional fit; a voice-matched generation step drafts proposals as DOCX in the Vice Chancellor’s own register.
//Role
Architect and engineer — ingestion, the scoring model, voice-matched generation, and the test suite.
//Outcome
Live in production with 2,043 grants ingested, 1,849 scored, 288/288 tests passing, and a delivered Vice-Chancellor demo.
What it took
Technical proof.
▸2,043 grants + 19 corporate programs ingested; 1,849 scored by Sonnet for institutional fit.
▸Voice-matched DOCX proposal drafts generated in the VC’s own writing style.