Junbo Wang

I am a researcher working at the intersection of multimodal artificial intelligence, generative modeling, and spatial intelligence. My research focuses on building AI systems that can understand, generate, and reason across diverse modalities, with an emphasis on learning from and modeling complex real-world environments.

Currently, I am a member of the GRIND Lab (Geospatial Responsible AI for Nature–Human Dynamics Lab) and the GISense Lab (Geospatial Intelligent Sensing and Mapping Lab). My work brings these AI directions into geospatial and environmental contexts, including natural hazards, disaster response, urban environments, and human-centered environmental perception. This interdisciplinary background motivates me to study how advances in artificial intelligence can be grounded in spatial, environmental, and real-world physical contexts.

Looking forward, I am interested in extending these ideas toward world models, multimodal and agentic AI, and embodied intelligence. In particular, I am interested in AI systems that can move beyond perception and generation toward richer reasoning, interaction, and eventually action in complex physical environments.

We introduce Geo-contextual Soundscape-to-Landscape (GeoS2L) generation, a task that emphasizes geographic consistency.

We systematically examine spatial explicitness in GeoAI and geospatial foundation models and provide a roadmap toward more spatially grounded artificial intelligence.

We use large language models to code maternal-health interviews for thematic analysis.

We meta-train node-specific transformations to enable graph few-shot class-incremental learning.