NexaHomes
Where AI finds your next home.
Visit live site
Overview
The challenge
Croatia's property listings are scattered across a dozen portals, stale, and impossible to compare. Buyers scroll for hours; agencies drown in manual data entry.
A basic search filter can't fix this — it can narrow a list, but it can't tell a buyer which of 200 near-identical apartments is actually worth touring. And manual curation by an agent doesn't scale past a handful of clients at a time.
The approach
Scrape
An AI pipeline continuously scrapes listings across sources and normalises them into one clean feed.
Rank
Every property is scored and ranked — surfacing the best matches, not just the newest.
Match
Buyers get an AI match score per listing; the front-end reads like a premium editorial catalogue.
Operate
A role-based admin panel runs the scraper, manages listings and monitors the whole system in real time.
Tech decisions
Listings needed to be server-rendered and indexable — buyers search "apartments Zagreb" on Google before they ever open an app, so pages had to be fast and crawlable from the first request.
Listings need real structure — price ranges, room counts, location filters that join cleanly. A relational database enforces that; a loose document store would let bad scraped data slip through.
A pipeline that continuously ingests messy third-party listing data needs strict typing to catch malformed records before they ever reach a buyer's screen.
Inside the build
Live walkthrough
01. Home Search
The whole experience, top to bottom — the AI search, the ranked listings, the match scores.
02. AI Engine
Watch the AI analyze inputs and rank properties based on custom match scores in real time.
03. Seamless Bookings
A flawless end-to-end booking flow designed to convert users with zero friction.
What it delivers
Unified feed from many sources
No more checking a dozen portals separately — every listing lives in one place, refreshed automatically.
Match scoring per listing
Buyers see which properties actually fit them, not just which ones were listed most recently.
Continuous auto-scraping
New listings appear on their own — nobody on the team has to manually re-check source sites.
Role-based access (user + admin)
Buyers and the operations team each get an interface built for what they actually need to do.