Back to Portfolio

OfferEngine - Real Estate AI Platform

Role: Developer

A fully automated real estate acquisition platform that processes 2 million+ properties end-to-end: from MLS listing ingestion to automatic offer generation with contracts attached. Built entirely by one developer in one year, this system leverages LLM-derived metrics for intelligent property valuation and automated decision-making. Technical foundation: FastAPI/SQLModel/Celery architecture processing data through a sophisticated pipeline with 7+ MLS integrations, OpenAI Batch API for LLM analysis, DocSpring for contract generation, and Salesmate CRM for deal management.

Project Screenshots
Offer Management Dashboard - Track automated offers across campaigns
Offer Management Dashboard - Track automated offers across campaigns
Campaign Dashboard - Real-time tracking of 2M+ properties across multiple markets
Campaign Dashboard - Real-time tracking of 2M+ properties across multiple markets
MLS Listing Processing - Raw data transformed into actionable intelligence
MLS Listing Processing - Raw data transformed into actionable intelligence
Property Analysis - LLM-derived ARV and repair estimates with comp analysis
Property Analysis - LLM-derived ARV and repair estimates with comp analysis
HCA Comparison Analysis - Cross-model LLM output normalization visualization
HCA Comparison Analysis - Cross-model LLM output normalization visualization
Campaign Configuration - ML-powered filtering with neural network scoring
Campaign Configuration - ML-powered filtering with neural network scoring
System Control Center - FastAPI/Celery stack with real-time task monitoring
System Control Center - FastAPI/Celery stack with real-time task monitoring

Impact & Results

Full automation: MLS listing → Delivered contract in 2 hours
2M+ properties analyzed with LLM-derived valuations
99.9% uptime with automated monitoring and self-recovery
Zero data loss through crashes - full idempotent design

Business Impact

Technical Architecture

Novel neural network for cross-LLM/cross-prompt normalization (potentially industry-first) 2TB+ PostgreSQL database optimized for sub-second queries at scale Complete CI/CD pipeline: GitHub Actions → GCP VM with automated deployments Idempotent architecture - every operation safe to retry, resistant to crashes Self-healing infrastructure with automatic recovery from failures FastAPI/SQLModel/Celery distributed architecture Zero-downtime deployments with health monitoring and automatic rollback Transient failure handling: retry queues, dead letter queues, error recovery OpenAI Batch API integration with retry logic DocSpring API for dynamic PDF contract generation Salesmate CRM integration with custom field mapping Secure production environment: SSH tunnels, GCP IAM, encrypted credentials Automated database migrations with Alembic version control Multi-environment support (local/test/prod) with environment-specific configs Docker Compose orchestration with health monitoring

Challenges & Solutions

Challenge 1

Automating entire flow: MLS listing → signed contract with zero human touch

Solution 1

End-to-end pipeline: MLS → LLM → Comps → ARV → Repairs → Margins → Contracts → Email

Challenge 2

Managing 2TB of calculated field data with sub-second query performance

Solution 2

PostgreSQL optimization for 2TB+ scale: partitioning, composite indexes, query optimization

Challenge 3

Building idempotent operations resilient to server restarts and transient failures

Solution 3

Idempotent task design with Celery - all operations safely resumable after crashes

Challenge 4

Incompatible scoring between LLMs (GPT-4 vs Claude) AND across different prompts

Solution 4

Novel PyTorch neural network mapping LLM outputs across models AND prompts (<100ms inference)

Challenge 5

Handling 7+ different MLS systems with wildly inconsistent data formats

Solution 5

Unified MLS provider abstraction layer with fallback sweep mechanisms

Challenge 6

Implementing zero-downtime deployments with rolling updates for 24/7 operation

Solution 6

Automatic retry logic with exponential backoff for transient failures

Challenge 7

Achieving 100% email deliverability for automated offers at scale

Solution 7

GitHub Actions CI/CD with smart build types ([app-only], [small], [optimized])

Challenge 8

Handling network partitions and database connection failures gracefully

Solution 8

Email reputation management with warming, rotation, and deliverability monitoring

Technology Stack

Python 3.12 FastAPI SQLModel Celery PostgreSQL Redis PyTorch Docker Compose Alembic GCP GitHub Actions Nginx SSL/TLS Supervisor