ChatWith.homes
AI-powered Real Estate Deal Finder
An AI-powered property intelligence system designed to help real estate investors, flippers, renovators, and developers discover and evaluate potentially attractive properties faster.
01 / CONTEXT
The problem
Real estate investors can spend hours manually reviewing listings, estimating renovation requirements, researching comparable properties, and calculating potential margins. Listing images and broker descriptions are also incomplete and inconsistent sources of property information.
02 / APPROACH
How I built it
Property listings and images are ingested and normalized. Vision models assess room-level condition, while local cost benchmarks inform renovation estimates. Financial models then calculate potential ARV, ROI, and profitability ranges.
03 / CONSTRAINTS
What had to hold
Low-quality property photos Unstructured broker descriptions Incomplete property information Uncertain renovation requirements Variable local material and labor costs Need for conservative financial projections
04 / OBJECTIVES
Definition of done
• Automate property discovery • Identify potential fixer-uppers • Estimate renovation scope • Model ARV and ROI • Research property context • Alert investors to matching opportunities
05 / THE SYSTEM
Architecture
Property ingestion Property feeds and associated images are processed by background workers. Geospatial intelligence Property coordinates and spatial information are stored using PostGIS. Multimodal analysis Property images are analyzed to estimate room-level condition. Financial modeling Renovation assumptions feed ARV, ROI, and deal-scoring calculations. Investor alert engine Matching opportunities are dispatched according to investor buy-box criteria.
06 / DECISIONS
Key decisions
Use multimodal property analysis Why: Images contain condition information that is not present in structured listing data. Combine AI analysis with cost benchmarks Why: Visual estimates need grounding in local renovation economics. Expose financial sensitivity Why: Real estate investors need to understand uncertainty rather than rely on a single optimistic projection.
07 / SHIPPED
Deliverables
• Automated deal discovery (in-progress) — Core capability of the product (see features). • Fixer-upper identification (in-progress) — Core capability of the product (see features). • Distressed-property discovery (in-progress) — Core capability of the product (see features). • Renovation scope estimation (in-progress) — Core capability of the product (see features). • Repair-cost forecasting (in-progress) — Core capability of the product (see features).
08 / LESSONS
What I'd keep
• Real estate underwriting needs conservative estimates. • Renovation estimates should expose sensitivity rather than imply false precision. • Multimodal AI becomes more useful when grounded in local cost and property data. • The goal is not simply to find more properties but to surface opportunities that match an investor's criteria.
What changed
Hours → seconds
Deal filtering and underwriting time
to date
// START A PROJECT
Building something with similar constraints?
If it's in your critical path and you'd rather not learn its failure modes live, let's talk about it.