Research
My research sits at the intersection of the built environment, urban data, and machine learning. Three threads, each with active publications:
Digital twins for the built environment
I build digital-twin representations of buildings and cities that combine BIM/GIS spatial structure with sensor streams and occupant feedback. The goal is to make spatial context first-class in machine-learning models of comfort, energy, and behaviour — not just a backdrop.
Representative work: Build2Vec (B&E 2022), What is a Digital Twin Anyway? (B&E 2025), The Internet-of-Buildings (J. Physics 2021).
GeoAI & street-view imagery
Cities have rich, public visual data — street-view imagery — that we’re only beginning to use computationally. My recent work at the NUS Urban Analytics Lab uses this data to study how urban perception varies across people and places, what makes greenery feel restorative, and how to acquire and process street view at scale.
Representative work: Global urban visual perception varies across demographics and personalities (Nature Cities 2025), ZenSVI (CEUS 2025), It’s not always greener on the other side (Landscape & Urban Planning 2026).
Graph neural networks for personal indoor environment
Indoor environmental quality (IEQ) is intensely personal — what’s comfortable for one person is not for another. I use graph neural networks over spatial-temporal data to predict individual occupant preferences and optimise sensor placement and feedback sampling.
Representative work: Personal thermal comfort models using digital twins (B&E 2022), Targeting occupant feedback using digital twins (B&E 2022), SpaceBrain (NUS Software Invention Disclosure 2021-140).
For the full programmatic vision — including the proposed CityBotics Lab — see the Research Statement →.
See the publications page for the full record, or the projects page for active and recent work.