Youssif Hegazy

He studied computer science and artificial intelligence at Benha University's Faculty of Computers and Artificial Intelligence, and has spent the three years since building production systems — the kind that keep running after the launch announcement, get handed to a client's own team, and have to survive a requirement nobody anticipated.

His work at the studio starts before any code does. He owns the architecture: the data models, the service boundaries, and the decisions that are cheap to make in week one and expensive to reverse in month nine. The recurring judgement is which parts of a system should be clever and which should be boring, and he is consistent about the answer — the interesting engineering goes into the AI layer, and everything underneath it is deliberately unremarkable so it can be reasoned about at three in the morning.

On the AI side he builds the things the studio sells: retrieval-augmented systems that answer from a client's own documents instead of guessing, conversational and voice agents that hold a real exchange in Arabic or English, and agentic systems that take actions across a company's tools rather than just describing them. The engineering problem in all three is rarely the model — it is retrieval quality, grounding, latency, and knowing what an agent should refuse to do.

The web and mobile practice runs on Next.js, React and TypeScript, with React Native and Flutter for cross-platform work and Swift and Kotlin where a native build earns its cost. He treats performance and accessibility as build constraints rather than a polish phase, which is the only way either of them survives a deadline.

He builds bilingually as a matter of course. Arabic is not a translation pass applied at the end of a project — it changes layout direction, typography, content modelling, and how an AI agent has to be evaluated, and designing for it from the first commit is considerably cheaper than retrofitting it.