Computer Vision Engineer
CS MSc student at TU Dresden, Germany. Interested in applying computer vision and machine learning in the medical field .
REE
Sep 2025 — Nov 2025Built a dataset of nearly 8,000 annotated basketball images for sports analytics model training.
• Basketball Video Dataset Engineering: Extracted video frames from recorded basketball matches, applied object detection models, and refined manual annotations to build a robust dataset of nearly 8000 images for sports analytics model training.
Oblix Tech
Aug 2025 — Sep 2025Benchmarked vision models on 360-degree construction imagery and validated zero-shot inference across six classes.
• Equirectangular Construction Site Analysis: Benchmarked DINOv2 and Grounding DINO and its finetuned variations for object detection and semantic segmentation in 360° construction imagery across 6 major classes. Validated that self-supervised models can support construction progress monitoring through zero-shot inference, reducing reliance on manual data annotation.
Cellula Technologies
Aug 2024 — Oct 2024Developed image classification and water segmentation pipelines, with the water segmentation model achieving 0.80 IoU.
• Oral Disease Classification with CNNs: Applied Convolutional Neural Networks including ResNet and VGG for image classification tasks using a custom-collected dataset and deployed with Streamlit, improving classification performance across 7 disease classes. • Water Segmentation: Engineered a water body segmentation pipeline using 12-channel multispectral and optical data with U-Net architectures, achieving 0.80 IoU.
Technocolabs Software
Aug 2024 — Oct 2024Trained mortgage payment prediction models and reported over 90% accuracy.
• Mortgage Payment Prediction: Trained machine learning models for classification and regression, applied feature engineering on a sparse dataset, achieving over 90% accuracy.
TU Dresden
2026Universität Passau
2026 — 2026Alexandria University
2020 — 2025Fine-tuned Qwen2, Mistral-7B, and BlenderBot-400M-distill on 6500 paragraphs from web-scraped data about Egyptian monuments and museums. Developed a full-stack Retrieval-Augmented Generation (RAG) system enabling users to query large Egyptian history documents using advanced AI. Integrated a React frontend with a Node.js/Python backend, Google Gemini 2.0 LLM, and embedding models via a persiste...
Developed an offline Arabic OCR system leveraging vision-language models (VLMs) for robust recognition of handwritten and printed text. Combined the KHATT handwritten dataset with a custom-augmented printed dataset to address script variability and to increase dataset size to approximately 12000 images. Applied advanced data augmentation and parameter-efficient fine tuning (PEFT) techniques, in...
