The Condo Book Project

Partial The search front end and database are standard, but turning thousands of unstructured NY AG offering-plan PDFs into reliable structured fields like budgets, reserves, and sponsor data is a hard extraction problem. Accuracy on legal documents matters for a buyer's decision.

What it would actually take: Build a pipeline that bulk-downloads plans from the NY AG filings source, parses PDFs with OCR where needed, and uses an LLM or NLP extraction with schema validation to pull financial and entity fields. Store results in a searchable database with a full-text and faceted search UI. The hard part is extraction quality and verification at scale, which benefits from someone with real-estate or legal document expertise to spot-check outputs.