High-Scale Property CRM Engine
Core production backend infrastructure behind Propforce, Zameen.com and Bayut.sa: DubizzleLabs' real-estate transaction CRM, serving over 40 million monthly users.
40M+ MAU · 45% lower query latency · 5,000+ mortgage clients
The system
A real-estate transaction CRM is deceptively heavy: high write volume, long-lived records with strict audit requirements, and read paths that fan out across listings, agents and clients. At 40M+ monthly users, the cost of a bad query plan is not theoretical. It is the difference between a page that loads and one that times out.
What I built
- Scalable backend services and REST APIs in Node.js, TypeScript and NestJS across the transaction domain.
- An end-to-end mortgage feature adopted by 5,000+ active clients, covering application through approval.
- RBAC-based logging and observability: metrics and tracing across every critical CRUD path, so failures were attributable.
- Parallel cron jobs and background workers that moved slow work off the request path and raised sustained throughput.
Performance work
The latency win came from unglamorous places: reshaping MySQL and PostgreSQL schemas around actual access patterns, adding the indexes the query planner was missing, and removing N+1 fan-out in the hot read paths. Core query latency fell 45%, which bought the platform headroom to keep growing on the same infrastructure.