Developing innovative data-driven and physics-informed neural network architectures for accurate ocean parameter forecasting. This project combines traditional oceanographic models with modern deep learning techniques to improve prediction accuracy.
Leading the development of advanced deep learning models for automated defect detection in urban infrastructure. Deployed in unmanned ground vehicles for real-time inspection of culverts and sewer systems with focus on handling data imbalance.
Pioneered explainable machine learning frameworks for human-AI collaboration in air traffic management. Created innovative trust measurement systems between AI and air traffic controllers to enhance joint decision-making capabilities.
Developed machine learning-guided failure detection techniques for advanced nanoscale semiconductors. Combined first-principles physics-based modeling with data-driven approaches for next-generation nanoelectronic designs.
Built evolving neuro-fuzzy-based intelligent control framework for flapping wing micro air vehicles. Developed novel autonomous learning algorithms addressing control challenges in highly dynamic aerial systems.
Developed novel Parsimonious Learning Machines addressing high-parameter bottlenecks in traditional fuzzy neural network designs. Created efficient algorithms for nonlinear system identification with reduced computational complexity.
Research on explainable few-shot learning techniques for remote sensing and computer vision applications. Developed attention mechanisms and feature aggregation networks for improved classification with limited training data.
Comprehensive review and research on ethical and robust large language models. Investigating safety, fairness, and reliability aspects of LLMs for responsible AI deployment.