Key research projects advancing trustworthy AI, bias mitigation, and medical AI systems for real-world impact.
Comprehensive open-source bias detection toolkit for medical AI with 4 fairness metrics: demographic parity, equalized odds, calibration, and disparate impact. Achieved 30.8% reduction in demographic bias while maintaining 98% model accuracy.
ML-based risk stratification system achieving 99% accuracy using XGBoost ensemble methods. Integrated homomorphic encryption for privacy-preserving computation on sensitive health data.
CheXNet-based diagnosis system for 14 thoracic diseases, integrating image and text modalities. Discovered 4.3× bias amplification when combining visual and textual medical data.
Multi-modal ECG analysis system supporting real-time monitoring, file upload, and paper ECG digitization. Implemented fairness-aware training achieving >0.9 fairness score while maintaining 95% accuracy.
Comprehensive framework integrating 4 disease models (Diabetes, Parkinson's, Heart Disease, Breast Cancer) with 92% average accuracy and real-time inference capabilities.
Comprehensive medical AI platform integrating computer vision, NLP, and predictive modeling. Implements bias-aware algorithms for fair diagnosis across demographic groups.
Specialized deep learning framework for medical imaging with custom architectures for handling small datasets and class imbalance common in medical research.
Comprehensive toolkit for medical image analysis including preprocessing, augmentation, and evaluation metrics specifically designed for healthcare imaging research.
Comprehensive statistical analysis package for medical research including power analysis, survival analysis, and meta-analysis tools with publication-ready visualizations.
Visit my GitHub profile to explore additional projects, code repositories, and open-source contributions.
View GitHub Profile