Structured 3D World Models
I develop expressive representations for structured physical environments that enable scalable learning, spatial reasoning and generative modeling in complex 3D domains. My work investigates geometry-aware encodings and latent field architectures as foundations for world models capable of generalization and synthesis across diverse spatial systems.
Surrogate Modeling for Climate and Environmental Performance
I build learning-based surrogate models that approximate high-fidelity physical simulation of climate-driven processes, including airflow, heat transfer and environmental field dynamics. This research advances scalable modeling of environmental performance under varying climatic conditions, enabling rapid prediction and analysis of complex physical systems.
Data as Infrastructure for (Geo)Spatial Intelligence
I treat data generation and curation as foundational infrastructure for spatial and geospatial AI. I develop simulation-driven and reconstruction-based pipelines that synthesize and align heterogeneous data across scales, from remote sensing and environmental fields to urban morphology and architectural form. By addressing structural data gaps and supervision limitations, I build physically grounded learning frameworks that enable robust and generalizable models for decision-making under real-world constraints.