AI Property Data Automation Platform

I need an AI platform/software that automates logging into local government websites, retrieves property records, and uses AI/OCR to extract and summarize key data — including matching records to the correct owner/property even when names or entries vary across sources. This is a paid first-phase engagement: build and validate the pipeline for a single test jurisdiction. Strong potential to grow into an ongoing role as the platform expands to additional jurisdictions.

Requirements

- Log into and interact with local government websites (varied, non-standardized interfaces, potentially dozens to hundreds of distinct portals over time) - Support structured search inputs (name, ID number, or address) - Retrieve property records including: Owner information, Property value, Transaction history, Liens, encumbrances, and other title-affecting documents. - Apply entity-resolution/matching logic to determine whether records across sources refer to the same person or property, using a confidence score rather than simple binary matching - Design storage using AWS S3 (documents) and a Postgres database (metadata), including caching previously retrieved documents to avoid repeat cost - Build human-in-the-loop functionality: flag uncertain or illegible items (e.g., poor handwriting, ambiguous dates) for user confirmation rather than guessing - Implement strong security/credential-handling practices, since the platform will hold login credentials for multiple external systems - Build an LLM-driven report generator that synthesizes extracted data into a clear, factual written summary Ideal Skills & Experience: - AI/ML development, including LLM integration - Web scraping and automation, specifically authenticated/session-based scraping (not just open APIs) - OCR/document-AI technologies, including messy or handwritten documents - Entity resolution / fuzzy matching with confidence scoring - Database and cloud storage architecture (AWS, Postgres) Looking for a tech-savvy developer who can deliver a robust, reliable first version and is interested in a longer-term relationship if this first phase goes well. Please include relevant past projects in your bid.