Data Doesn't Lie,
People Do.

Built an end-to-end RAG (Retrieval-Augmented Generation) system for e& Egypt to assist new employees during onboarding. The system allows the company to upload internal documents, policies, and training materials, enabling new hires to quickly get answers to common questions without manual search.

• Built for e& Egypt onboarding — helps new employees find information quickly
• Document management system with intelligent Q&A interface
• Vector database with sub-second response + semantic search capabilities
• Reduced HR workload by automating common onboarding questions

Advanced ML pipeline achieving 91.59% AUC using XGBoost for predicting customer churn across 2.58M+ telecom customers in Mauritania and Senegal.

• Evaluated 5 ML algorithms (XGBoost, LightGBM, CatBoost, Random Forest, Logistic Regression)
• Custom preprocessing transformers handling 140+ unique categorical levels
• Deployed real-time churn prediction Streamlit app

End-to-end ML platform for diabetes prediction achieving 87.66% accuracy through ensemble learning.

• Two-stage training pipeline with MLflow experiment tracking
• Streamlit dashboard with 7 interactive modules including XAI analysis
• Advanced feature engineering pipeline creating 16 derived features

Increased marketing campaign effectiveness by 30% through behavioral segmentation of 8,950+ credit card customers into 4 distinct groups using K-means clustering.

• Achieved 94% clustering accuracy through comprehensive preprocessing
• Reduced customer acquisition cost by 20%
• Delivered clear visualizations and insights for business stakeholders

IoT-based accident detection and alert system that automatically detects vehicle crashes and notifies emergency services.

• Real-time sensor fusion for accurate crash detection
• GPS tracking and automatic emergency notification
• 1st Place at IEEE Mansoura Victories III (70 teams)
• Top 6 Finalist at Hack'24 — AUC (141 projects)