For my Digital Media Engineering degree at German University in Cairo, I built a two-stage aerial object recognition pipeline for DOTA. Swin-Tiny with 8-view test-time augmentation reached 97.20% accuracy.
I also built a Graph-RAG travel assistant connecting a Neo4j knowledge graph to Gemma, Llama, and Mistral. It uses Cypher queries, vector similarity search, and intent classification to ground answers in graph knowledge.
For brain MRI tumor segmentation, I compared a baseline CNN with a U-Net using TensorFlow/Keras. The U-Net achieved a 0.66 Dice score on tumor slices, compared with 0.50 for the baseline.
At Al Nour Association, I helped build a live e-commerce store and developed its product catalog and category pages as a Full-Stack Developer Intern. I also supported Python lab sessions as a Junior Teaching Assistant at German University in Cairo.

