Building systems that survive real operations

I am an AI Product Engineer and hands-on Tech Lead at Pixelity, based in the UAE. Over seven years I have shipped production software across real estate, retail, F&B, commerce, service operations, and space systems, often owning the work end to end.

The last three years have centered on applied AI: LLM pipelines for document intelligence and structured extraction, retrieval systems with hybrid search and reranking, agent workflows, and the validation layers that make model outputs dependable in real operations.

I take messy business workflows, understand the people and data around them, and ship the smallest reliable system. Current work is led by RayaOS, an operating platform for real estate brokerages that connects CRM, leads, listings, deals, documents, rentals, and governed AI.

How a system behaves under pressure matters as much as what it does on a good day. I treat a model’s confidence and its permission to act as separate questions, build autonomy in stages rather than a single on/off switch, and keep a control that stops any automated action immediately, per tenant, without a deploy. Documentation gets the same discipline: what a system is meant to do stays separate from what has actually been verified, so nobody inherits a plan dressed up as a status report.