At Seha Care, I develop Python and FastAPI services for clinical evidence extraction, normalization, and downstream triage. I also designed Transformer-based components for Arabic dialect classification and clinical text normalization.
I built reproducible benchmarking pipelines for Transformer models, Arabic LLMs, and ASR systems, using gold-standard datasets and automated metrics. I also contributed to the architecture of a Medical Knowledge Graph and deterministic clinical decision engine.
At Zaia Systems, I designed a secure multi-tenant backend with JWT, OAuth2, RBAC, encrypted credential storage, and tenant-level data isolation. I also built email automation workflows and improved reliability through caching, database indexing, and integration testing.
In open source, I implemented a Median Filter node for Graphite Editor using Quickselect, and contributed fixes and usability improvements to the OpenStreetMap iD Editor. I also built a license-plate recognition and access-control project that achieved 96% detection accuracy on Arabic plates.
