Spatiotemporal Intelligence
We model how complex environments evolve over time and space, with an emphasis on forecasting and generalization.
- Video and sequence prediction
- Weather and satellite intelligence
- Urban dynamics and time-series modeling
Our Research
Robust and generalizable AI for complex real-world challenges.
Our research is dedicated to AI for Social Good, grounded in computer vision and machine learning. We pursue interdisciplinary methodology that connects perception, prediction, and multimodal reasoning to high-impact real-world problems.
Vision & Multimodal AI
We study visual and multimodal understanding that can remain useful across changing tasks, domains, and observations.
- Computer vision and video understanding
- Vision-language models and multimodal reasoning
- Domain adaptation and generalization
AI for Social Good
We connect methodological AI advances to societal challenges where reliable prediction and understanding can create public value.
- Climate, weather, and disaster-related AI
- Urban context modeling and public safety
- Human-centered intelligent systems
Physical & Embodied AI
We explore intelligent systems that perceive and reason about human, industrial, and robot-centered physical environments.
- Environment-aware prediction
- Industrial anomaly understanding
- Robot and human-centered perception