Omar Shehab

Refactoring the backend by migrating from JavaScript to TypeScript, optimizing database queries and data flow to enhance system performance.

• Designed and implemented a robust machine learning system for materials analysis, optimizing data processing by clustering over 264k Metal-Organic Frameworks (MOFs) using the Leiden algorithm.
• Validated cluster quality through a comprehensive multi-metric evaluation, ensuring high reliability in data interpretation.
• Developed a high-performance LightGBM classifier achieving 90% accuracy in predicting molecular clusters, significantly improving from a 50% baseline.
• Enhanced model performance by addressing data imbalance through SMOTE and leveraging K-Means for effective spatial feature engineering.
• Gained hands-on experience in backend development and database management, reinforcing skills in software engineering and AI integration.

• Engineered a Retrieval-Augmented Generation (RAG) system for YouTube videos, facilitating enhanced semantic navigation through transcripts.
• Architected an efficient three-tier caching system (L1 in-memory, L2 Redis, L3 Postgres), achieving an impressive 95.8% cache hit rate, optimizing data retrieval.
• Developed a robust streaming pipeline to process 1-hour videos in approximately 157 seconds, utilizing 30-second chunks and executing 4 parallel jobs for improved performance.
• Achieved a significant reduction in processing time by over 65%, enhancing overall system efficiency and user experience.
• Leveraged skills in software engineering, backend development, and database management to deliver high-quality solutions.