Manzel Masr · an Arabic-first property marketplace with a broker CRM built in

The problem#

Buying or renting a home in Egypt meant scrolling Facebook groups, calling brokers who might not answer, and having no way to compare listings on a map with real prices. Brokers, on their side, tracked leads in notebooks and chat threads and lost them.

What I built#

• A marketplace for buyers and renters, Arabic-first and right-to-left: map-based browsing with live prices, filters for price, type, location, rooms and area, and a "property reels" feed of short video tours.
• Direct WhatsApp-to-broker messaging from every listing — the channel Egyptians actually use — with verified broker and developer directories so people know who they're talking to.
• A 24/7 AI assistant that searches listings conversationally: it turns a question like "a two-bedroom flat in 6th of October under four million" into a real search, using retrieval over the live database (pgvector, LangChain, OpenAI).
• A CRM for brokers: their listings, their leads, and the follow-ups, in one place, on their phone.
• Mobile apps for Android and iOS on top of the same backend.

How#

Django and PostgreSQL at the core (with pgvector for the AI search), a React frontend, the WhatsApp Business API for messaging, and OpenAI models behind the assistant. Founder-led, from the first data model through launch, the sales conversations with brokers and the daily running of the product.

Outcome#

A live product in a market that had nothing like it, with real listings across Cairo's new cities, an assistant that answers in Egyptian Arabic, and a broker CRM that brings the supply side onto the platform. It's also my daily reminder of what founders need from an engineer: working software, honest scoping, and no surprises.