REI Engine
PostGIS + rules engineA real-estate intelligence platform with two halves: a geospatial ingestion pipeline that pulls public county property data into PostGIS, and a financing-eligibility rules engine that ranks how to actually fund a given property across federal and state programs. A domain I learned by building it.
The Problem
The data exists. It is scattered and unusable.
County parcel and distress data lives behind dozens of inconsistent government GIS endpoints, and the question that actually matters, how would I finance this specific property, takes an expert to answer. Both problems are tractable with the right ingestion and rules layers.
The Build
Ingestion plus a financing rules engine.
A config-driven ArcGIS adapter normalizes parcel data into a PostGIS property graph with idempotent, dedup-by-APN upserts. On top of it, an effective-dated rules engine evaluates a property and a borrower profile against program criteria and ranks viable financing strategies, with stacking and graceful degradation.
What makes it interesting.
Idempotent geospatial ingestion
Config-driven ArcGIS field maps, pagination, and a source-hash short-circuit. Re-running the ingest produces zero duplicates.
Three-valued logic engine
A pure Kleene interpreter returns eligible, ineligible, or needs-input, so the system asks the exact missing question instead of guessing.
Effective-dated rules
Program criteria are versioned with GiST no-overlap constraints, so eligibility is always evaluated against the rules in force at a given date.
Tech stack.
Ingestion
- Python
- ArcGIS REST
- PostGIS
- Idempotent upserts
- Source-hash dedup
Rules Engine
- Kleene three-valued logic
- Effective-dated criteria
- GiST exclusion constraints
- Strategy stacking
API & Dev
- FastAPI
- PostgreSQL
- docker-compose
- pytest
- Next.js dashboard (planned)
Verified.
Core ingestion and financing engine shipped and verified locally. Dashboard UI and multi-county expansion in progress.