AI Systems Engineer & CS (AI) student in Cairo, building AI systems that hold up outside the notebook — from LLM agent pipelines to deployed computer vision models.
I'm currently working as an AI Systems Engineer, building end-to-end LLM-agent automation pipelines integrated with CRM systems, and previously interned as an API Engineer designing and securing REST/SOAP APIs on GCP with Apigee X.
CNN-based deepfake detector trained on FaceForensics++ — 90.4% accuracy, 0.96 AUC
- Face detection with RetinaFace + EfficientNetB0 two-phase transfer learning
- Grad-CAM visualization to interpret model decisions
EfficientNetB0RetinaFaceTensorFlowGrad-CAMFaceForensics++
huggingface/transformers ⭐ 162k
PR merged into one of the most-starred repos on GitHub
- Identified and fixed a documentation bug where SmolLM3 was miscategorized under Multimodal models
- Reviewed and approved by co-founder Julien Chaumond, merged into
huggingface:main
