// work / software development

Property listings platform and public API

A real-estate listings platform with a public search API, an admin CMS for agents, geographic search and an automated image pipeline.

Client
Property marketing startup
Industry
Real estate
Year
2025
Role
Back-end lead and API design
Duration
14 weeks
Modern apartment building with glass balconies at dusk
Photo: @tobiasw on Unsplash

Context

A property marketing startup had a listings website that worked well enough for visitors but could not be reused: two partner portals wanted the same listings, and the founders wanted to sell access as a product. The existing site had the data model tangled into the page templates.

The problem

Build a proper platform: a public API that the startup’s own site and its partners would all consume, a CMS for agents to manage listings, and fast geographic search. The migration had to keep the existing site live and its search rankings intact.

Approach

API first. I designed the API before touching any screen: listings, agents, media, saved searches and webhooks, documented in OpenAPI and reviewed with the first partner before it was built. The startup’s own Next.js site became just another API client, which proved the API was good enough to sell.

Search. Listings live in PostgreSQL with PostGIS for location. Radius, polygon and “near this station” searches are a single indexed query. Facets such as price bands and bedroom counts are precomputed on write, so the search page renders from one request. Redis caches the hottest searches for a few minutes.

Images. Agents upload whatever they have, often 8 MB phone photos. A queue-backed pipeline validates, strips metadata, resizes to four sizes and generates WebP, then marks the listing as publishable. The CMS shows each step so agents are never guessing why a photo has not appeared.

Keys and metering. Every partner gets an API key with per-minute limits and a usage dashboard. Metering is written to a separate table and rolled up hourly, so billing can be built on it later without touching the request path.

The old site was redirected URL by URL to the new one, with a 301 map generated from the database, so nothing dropped out of search engines.

Outcome

The platform launched with the startup’s own site and three partner portals on the API. Search responds in around 120 ms at the 95th percentile with 40,000 listings. The image pipeline removed the single most common support request (“my photos aren’t showing”). The founders now have an API product with metering, which was the point.

Need something like this?

This is the kind of work I do under software development. Tell me about your version of the problem.