I've built and deployed production AI systems spanning Arabic legal research, medical imaging, satellite segmentation, and real-time computer vision. My Arabic Legal Assistant retrieves Egyptian legal articles with mechanically verified citations to prevent hallucinated references from reaching users.
At Arias Tech Solutions, I train and deploy AI products including multi-species sea turtle detection, a RAG-powered sea turtle chatbot, and an oil-spill segmentation pipeline that achieved 0.94 DSC on an external held-out test set. I integrate models into public-facing AWS applications and production FastAPI APIs.
As a freelance AI engineer, I built a clinical imaging platform validated by four practicing radiologists, who reported reduced diagnostic time and fewer missed abnormalities. I deployed a 3D UNet for hippocampus segmentation at 0.90 DSC and an EfficientNet pneumonia classifier at 0.90 AUROC.
I enjoy taking models from data preparation and evaluation through reliable deployment, using Python, PyTorch, FastAPI, Docker, cloud infrastructure, and modern agent frameworks.
