Yasmina RHARI

At I-Gouvernancia, I developed a fully local medical voice assistant for five languages, including Moroccan Darija and Amazigh. Its pipeline connects speech recognition, translation, Qwen2.5, and Arabic text-to-speech without sending patient data off-site.

I also assembled a fine-tuning corpus of 15,288 sentences and a 549-entry correction dictionary for under-resourced languages. For Amazigh, I compared four translation approaches.

I designed a clinical prioritization engine that scores urgency using patient age, medical history, recent camera events, and machine-learning confidence. It distinguishes between patients with identical symptoms where the previous indicator did not.

On NeroviaFlow, I connected an ESP32 water meter to a dashboard for leak detection and alerts; the project reached the national Hult Prize final. I’ve also developed a real-estate management application at OMA Investments Limited and an e-commerce application at SO SOLUTION.