Skip to main content
AI/MLJul 2020

Home Rumble

AI-powered property discovery platform that reinvents house hunting with machine learning-driven matching, collaborative search, and behavioral analytics - built on a fully serverless AWS microservices architecture.

Home Rumble - House Hunting Reinvented, AI-powered property discovery platform
Home Rumble - House Hunting Reinvented, AI-powered property discovery platform

The Challenge

Traditional real estate platforms rely on basic keyword filters - bedrooms, price, zip code - forcing buyers to manually sift through hundreds of irrelevant listings. Nuanced preferences like neighborhood feel, commute patterns, or natural light are impossible to express. The result: buyers spend weeks searching, agents waste time on poor-fit showings, and great properties go undiscovered because they don't match rigid filter criteria.

Our Solution

We architected a fully serverless property discovery platform on AWS. Elasticsearch handles sub-second full-text and geospatial search across millions of listings. AWS SageMaker ML models analyze user behavior - saved searches, time spent on listings, swipe patterns - to build preference profiles and surface increasingly personalized recommendations. The frontend is a React SPA with infinite scroll, map-based browsing, and collaborative boards where families can save, comment on, and rank properties together. ECS Fargate containers handle data ingestion from MLS feeds, while DynamoDB provides low-latency reads for user sessions and preference data.

Results & Impact

  • Reduced average property search time from weeks to minutes
  • Scaled to handle 10K+ concurrent users with zero downtime
  • ML-driven matching outperformed keyword search by 40% in user satisfaction
  • Microservices architecture enabled independent scaling of search, ML, and ingestion services
  • Collaborative boards increased household engagement by 3x vs solo browsing

Technologies Used

AWS ECSAWS FargateElasticsearchAWS SageMakerDynamoDBReact

Services Used

Client Testimonial

HalSoft architected our ML-powered property matching from scratch - it outperformed keyword search by 40% in user satisfaction and scaled to 10K concurrent users with zero downtime.

Vazgen Avakyan

CTO, Avakyan Capital

Interested in a Similar Project?

Let's discuss how we can build something great together.