[← back to projects]

[live]

County Lead Intelligence

Real-estate lead engine built on public county records

3 months

A daily pipeline that turns scattered county court and recorder filings into owner-resolved, equity-scored leads for real-estate investors.

County Lead Intelligence screenshot 1
County Lead Intelligence screenshot 2

The problem

Investors look for motivated sellers — pre-foreclosures, probate, evictions, tax delinquencies — but those signals are buried in dozens of incompatible county recorder, court, and assessor systems, many behind captchas, logins, or bot protection.

What I built

A per-client lead platform: scrapers for each county run every morning, resolve each filing to a parcel and owner, estimate equity, skip-trace contact details, and push the results into a filterable dashboard and the client's CRM. I was the lead backend engineer across deployments covering ten states.

Core features

  • -Daily scheduled scrapes per county with live progress streaming to the dashboard
  • -Filing → parcel → owner resolution through assessor and GIS parcel layers
  • -Estimated equity, replaced by real loan balances when a PropStream export is imported
  • -Automatic skip tracing with per-batch spend caps
  • -Idempotent CRM push that never creates duplicate records on retry
  • -Lead table with facet filters, zip filtering, list stacking, and CSV export

Engineering highlights

  • -Turned an ASP.NET recorder search that "needed a real browser" into a plain HTTP scraper by reproducing its hidden date-picker state
  • -When a county index couldn't place a filing, read the recorded document itself — OCR on scanned TIFFs or the PDF text layer — and cross-checked the result, because a wrong address is worse than a missing one
  • -Transport ladder (direct → scraping API → residential proxy) with challenge-page detection so a blocked response never counts as success
  • -Per-source scrape trace (fetched → parsed → kept → reasons) persisted to the database, so "why did this county return 0 leads?" has an answer
  • -Equity model from assessor value and historical mortgage rates, amortized to today and labelled as an estimate

Tech stack

TypeScriptNode.jsExpressPostgreSQLnode-croncheerioTesseract OCRZodVitestReactViteTailwind CSSZustandRailwayVercel

Impact

  • -Going direct to one county's recorder surfaced 27 trustee sales in two weeks, where the previous source had produced 1 lead for the whole month
  • -Scrapers live across ten states, each deployment tailored to its counties and lead types
  • -25-file test suite on the main deployment

Want something similar built?

Happy to talk it through — drop me a line.

[get in touch]