I work across the parts of AI systems that tend to be underspecified: retrieval that has to be defensible, agents that reason over structure rather than flat text, models that have to stay interpretable because a clinician will act on the output.

Final-year Master's and engineering student in AI & Data Science at ESI Sidi Bel Abbès, graduating 2026. Research intern at TÉLUQ Université, Québec, working on early prediction of hypoglycemia from continuous glucose monitoring (CGM) — the constraints are interpretability, generalisation across patients, and running on resource-limited devices, because a clinician has to act on the output. Paper drafting targets the end of Q3 2026.

On the engineering side: I co-founded Novera, a dev agency, and the work there is less research and more consequences — a production e-commerce platform for a parapharmacy, national delivery integrations, a real admin dashboard that someone actually uses every day. That gap between a working prototype and something a non-technical client can operate at 11pm is where I spend a lot of time.

Before all of this: trained operators in cybersecurity at CNFE El Bayadh, administered databases at CNAS, and ran a small clothing shop called Tisharwatt for five years, which taught me more about latency and user patience than most of my coursework did.

Stack.