Private Markets Intelligence
Turns a firm's deal history and documents into a queryable intelligence asset
Overview
I designed a system that turns a private markets firm's scattered deal history and documents into a single queryable intelligence asset that compounds over time. I delivered the full architecture and a screen-by-screen product spec, from bulk document ingestion to a cited conversational interface. I also built in an honest coverage model, because a research tool that hides what it does not know is dangerous.
The problem
A firm's deal knowledge is buried across years of documents and deal history, so it cannot be searched, cited, or reused when the next decision comes up.
What I built
I architected bulk document ingestion, a cited conversational ask interface, confidence-scored auto-classification with human review, and a coverage model that surfaces missing data, all on Postgres with row-level security.