cv

Research Assistant – Knowledge and Data Engineering Lab

University of Tsukuba, Japan

Developed pipelines for transforming unstructured text into structured knowledge graphs using LLMs, improving retrieval efficiency by 40%.

• Built scalable ETL pipelines for PDF-to-graph transformation.
• Optimized LLM parsing strategies to standardize diverse recipe formats.

Undergraduate Research Assistant – NeuroInformatics Lab

University of Tsukuba, Japan

Analyzed EEG data to improve sleep stage classification using deep learning and generative data augmentation techniques.

• Achieved 75% F1 score with Transformer-based models on EEG signals.
• Used DeltaEncoder and cGAN for balancing highly imbalanced sleep data.

Data Science Intern

Dell Technologies, Egypt

Developed CNN models for dementia MRI classification and applied Grad-CAM for model interpretability.

• Achieved 92% F1 Score on multiclass dementia classification task.
• Highlighted key brain regions using Explainable AI (Grad-CAM).

Public Cloud Intern

Orange Business, Egypt

Built cloud-native deployments using Docker, Nginx, and Kubernetes with load balancing and HA configurations.

• Awarded Best Internship Project.
• Deployed full-stack application to Kubernetes with public Ingress.

Exchange/Research Program

University of Tsukuba

NeuroInformatics and Data Engineering

Bachelor of Science

Egypt-Japan University of Science and Technology (E-JUST)

Computer Science (Artificial Intelligence and Data Science)

• Probabilty and Statistics
• Linear Algebra
• Discrete Mathematics
• Calculus
• Data Science
• Data Mining
• Computer Vision
• Machine Learning
• Natural Language Processing

Best Internship Project

Orange Business

Awarded for outstanding cloud deployment project integrating Docker, Kubernetes, and public cloud architectures.

Best Presenter

Orange Business

Recognized for excellent technical presentation and communication skills during Orange Business internship.

Fashion Products Chatbot

Semantic search chatbot for fashion product recommendation using FAISS, LangChain, and Hugging Face Transformers.

• Deployed a FastAPI service for product retrieval.
• Embedded and indexed fashion product descriptions.

Real-Time ASL GPT

Built a real-time American Sign Language recognition system integrated with Retrieval-Augmented Generation (RAG).

• Trained SqueezeFormer model on 200GB ASL video data.
• Real-time low-latency gesture recognition pipeline.

EEG Sleep Classification

Transformer-based deep learning model for EEG sleep stage classification, using cGAN and DeltaEncoder for data augmentation.

• Addressed class imbalance using generative models.
• Achieved 75% F1 score across multiple sleep stages.

Alzheimer-XAI

Explainable AI project classifying dementia stages using CNNs with Grad-CAM visualization for clinical interpretability.

• Achieved 92% F1 Score in dementia classification.
• Generated Grad-CAM maps highlighting critical brain regions.

Sparse Supervision for Remote Sensing Segmentation

DeepLabV3+ based semantic segmentation for satellite images with only 1% labeled data using Partial Cross-Entropy Loss.

• Achieved 0.43 mIoU under extreme label sparsity.
• Evaluated different label densities (1%, 5%, 10%).

Knowledge Graphs using LLM

Automated extraction and transformation of unstructured text (recipes) into knowledge graphs using LLMs like GPT-4 and Mistral.

• Built end-to-end ETL pipelines reducing processing time by 40%.
• Standardized and structured 1000+ documents into graph format.