Featured Projects
Parkinson's Disease Progression Prediction
Analyzed 980K+ protein and peptide records to model early-stage Parkinson's progression using statistical and ML techniques. Applied advanced feature engineering and signal processing to achieve 65–70% classification accuracy on real patient datasets. Identified key protein biomarkers that improved interpretability and clinical applicability of predictive models.
AI-Powered Crop Monitoring for Precision Agriculture
Partnered with Dubai Future Labs to deploy drone-based crop analytics, optimizing agricultural resource allocation and sustainability. Processed 10,000+ aerial images and trained 3+ ML models to classify crop health with high accuracy and reliability. Integrated AI outputs into a responsive web dashboard adopted by field partners; recognized as Best Presentation at ICRAE 2023. Published findings in Springer Nature's DCAI 2024 conference proceedings, expanding academic and industry visibility. View Publication
Sign2Sign 3D Avatar Translator
Developed a 3D avatar-based English–German sign translator using vision and NLP models to enhance accessibility for deaf and hard-of-hearing users. Curated and labeled 5,000+ sign images to strengthen model robustness and real-world recognition accuracy. Trained and deployed dual-direction models (sign-to-text, text-to-sign) using Euler angle wrist motion calculations for precise avatar movement.
Intelligent Class Scheduling Optimization
Developed a genetic algorithm-based scheduling system that reduced class scheduling time from days to 2-3 hours while maintaining 90% accuracy. Optimized resource allocation and conflict resolution for large-scale academic institutions.