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.