Holofy · Case study
One search across 20+ property portals.
A renter describes a home in their own language. Holofy checks live listings, ranks them against that brief and puts the matches on one map.

- Industry
- Real estate
- Reach
- 9 countries
- Project
- SOME1 · 2025
The search behind the tabs
A buyer or renter can repeat the same search on several portals and still miss a relevant home. Countries and portals describe listings differently. A person looking for a place near a park has to translate that preference into whatever filters each site offers.
Holofy’s brief was to interpret the search in the person’s own language, find real listings from the relevant local portals and present the results on one map. Work, Meet AI’s founder built the project at SOME1.
A conversation, then a map
A shopper describes their preferred location, budget and the sort of home they want. The search separates factual limits from descriptive preferences. It queries live portals including Rightmove, Zoopla, idealista and Bayut, then ranks the matches for fit.
The map adds points of interest around a listing. Buyers and renters can watch available listing videos, save favourites and ask a voice agent to call an agency to arrange a viewing.
What the project covered
- 9 countries
- The original case reports market coverage across the UK, the UAE and Europe.
- 20+ portals
- The platform retrieves property listings from more than twenty live sources.
- 0 paid placements
- The documented search ranks listings by how well they fit the request.
- Local formats
- The system parses listings in multiple languages and supports different currencies.
- Agents and agencies
- Buyer accounts and saved homes sit alongside agency onboarding and voice-assisted viewing bookings.
How a request finds a listing
The query layer turns free text into filters and an embedding. A hybrid search combines SQL with semantic matching in Supabase and pgvector. Each portal has its own ingestion module, with Firecrawl and Playwright collecting live data.
A tool-using agent can search, geocode, handle currencies and find agencies. Mapbox GL and deck.gl display the results. Vapi and Twilio handle voice calls to agencies.
- Next.js 16
- React 19
- TypeScript
- Supabase + pgvector
- OpenRouter
- GPT-4o-mini
- Perplexity Sonar
- Firecrawl
- Playwright
- Mapbox GL + deck.gl
- Vapi + Twilio
Explore the original project
The figures and product interface come from SOME1’s public Holofy case study.
Visit Holofy
