This project aims to develop an integrated, physics-informed AI framework combining transformer-based field inference methods with adaptive parameter estimation from sparse sensor data.

In this project, we have developed scalable urban air pollution monitoring systems that combine physics-informed models, machine learning, and multi-modal sensing to detect and analyze pollution hotspots. By integrating image-based AI models and sparse sensor networks, we provide fine-grained, real-time insights into air quality across densely populated cities.

This project addresses urban traffic congestion by first characterizing the emergence of sudden traffic jams using coarse-grained speed and flow data, and then mitigating them through reinforcement learning-based speed control that requires no infrastructure changes. Together, these approaches enable a scalable, physics-informed framework for proactive and decentralized traffic management.

Investigating data processing pipelines and inference algorithms for single cell RNA sequencing datasets.