I'm Ahmed, an AI engineer. I build LLM systems end to end — retrieval, agents, fine-tuned models and the services around them — and I treat every change as an experiment: define the metric, measure the baseline, change one thing, measure again. No demo-ware.
My route into AI was unusual. I started in quantum computing — as a remote research intern with QUAE (QWorld's UAE chapter) I built an EEG classification pipeline with a Quantum Variational Autoencoder, which became a published paper at the Neurogenomics 2024 conference. That experience taught me how to read papers, run rigorous experiments, and ship reproducible work.
Now I focus on what makes an LLM system dependable in production: evaluation that catches regressions, tracing that shows why an answer went wrong, guardrails tested against hostile input, and the latency and cost of every call. In my latest project a prompt change is only promoted after it beats the current version on the same tests, and every result is written down in an improvement log.
I like working across disciplines — from generative vision and quantum computing to language models. As a Computer Science teaching assistant I designed and led 50 Compiler Theory lab sessions for 600+ students and led tutorials in Linear Algebra, Calculus and Information Retrieval, which sharpened how I explain complex ideas simply.
