Byeonghwa Jeong

I am Byeonghwa Jeong, an R&D Associate at Oak Ridge National Laboratory and a geographer and spatial data scientist. I develop computational and spatial analytical methods - including deep learning, machine learning, graph neural networks, computer vision, remote sensing, and spatial statistics - for population and socioeconomic measurement, urban change, disaster impacts, and infrastructure risk.

At ORNL, I contribute to LandScan Global and LandScan Mosaic and develop GeoAI methods for large-scale population modelling. My current work includes multi-scale population forecasting with graph neural networks, population-surface reconstruction using AlphaEarth satellite embeddings, building-level wealth estimation, and hurricane-induced power-outage forecasting. I also work with large-scale GPU/HPC workflows to develop geographically transferable models for complex spatial datasets.

I received my Ph.D. in Geography (Geospatial Data Science) from University College London, supervised by Professor Paul Longley and Professor James Cheshire. My doctoral research developed spatial and deep-learning approaches to study retail gentrification and high-street change. I later worked as a Postdoctoral Research Fellow at the University of Toronto School of Cities, where I developed computational methods for downtown recovery, building-level analysis, footfall data, and explainable machine learning.

My broader research goal is to develop scalable, interpretable, and geographically transferable geospatial AI methods that support real-world decision-making. I am particularly interested in linking population dynamics, socioeconomic conditions, the built environment, and infrastructure risk through spatial data science.