Research Statement

My research sits at the convergence of artificial intelligence, digital twins, and built-environment science. Over sixteen years I have moved deliberately from computational fluid dynamics and building simulation, through machine learning and graph neural networks, to urban-scale AI and multi-agent systems. The unifying thread is a commitment to making buildings and cities more responsive, efficient, and human-centred through data and intelligence.

Foundations: from simulation to spatial AI

My MSc at Mansoura University established deep expertise in Computational Fluid Dynamics (CFD) applied to architectural and urban design, using ANSYS and OpenFOAM to optimise natural ventilation in Egyptian intermediate cities. This work produced publications in Urban Climate, Building Simulation, and Frontiers of Architectural Research, and grounded me in the physics of the built environment that underpins all subsequent AI-driven work.

In parallel, I developed two practitioner-facing software tools that moved AI from theory into architectural workflows: ANT (machine-learning plugin for Rhino/Grasshopper) and GhCPython (a bridge enabling high-performance scientific computing libraries — scipy, numpy, scikit-learn — inside parametric design environments). These tools were adopted by practitioners and taught in computational design workshops, demonstrating my commitment to translating research into usable instruments.

My PhD at NUS developed Build2Vec and SpaceBrain — a framework that encodes building spatial data from BIM into graph structures and feeds them into Graph Neural Networks (GNNs) to predict occupants’ indoor environmental satisfaction with 14–28% accuracy improvement over conventional baselines. This was among the first applications of GNNs to building-scale occupant behaviour, and it produced five publications including two in Building and Environment and one in Energy and Buildings.

Current work: urban AI and geospatial intelligence

My postdoctoral work at the NUS Urban Analytics Lab (2023–2025) expanded the spatial-AI paradigm from building to city scale. Three active research threads:

Street-view imagery analytics (SVI). Co-developed ZenSVI, an open-source platform for large-scale acquisition, processing, and analysis of street-view imagery. Applied to global studies of urban visual perception, greenery, and demographic variation — published in Nature Cities and Landscape and Urban Planning.

Digital twin definition and methodology. Led a systematic analysis of 15,000+ scientific publications to derive a rigorous, data-driven definition of digital twins for the built environment — published in Building and Environment (2025).

Crowdsourced urban data analytics. Developed pipelines that extract and analyse behavioural and perceptual data from Google Reviews and other crowdsourced platforms to understand human experience of urban space at scale.

Future programme: CityBotics Lab

My next research programme — CityBotics Lab — is dedicated to multi-scale agentic digital twins that couple physical AI agents with city-scale simulation. The core idea is to convert any physical asset (building component, construction equipment, infrastructure node, public-space element) into an intelligent, autonomous AI agent embedded within a living, multi-resolution digital twin of the urban environment. This architecture enables real-time optimisation of building performance, construction processes, and city operations simultaneously.

Initial research thrusts:

Agentic building digital twins. AI agents embedded in building systems (HVAC, façade, occupancy) that negotiate in real time to minimise energy use while maximising occupant comfort — particularly relevant in extreme climates.

Urban environmental sensing at scale. Combining SVI, satellite imagery, IoT networks, and crowdsourced data to generate real-time digital twins of cities, supporting urban-heat-island mitigation and net-zero urban planning.

The lab is designed to attract funding from national and international research agencies (NSF, EU Horizon, UKRI, and regional equivalents) and to integrate undergraduate and graduate students from day one — building a pipeline of researchers trained at the intersection of AI, robotics, and built-environment science.

Metrics and trajectory

To date my work has produced 18 indexed publications in journals including Nature Cities, Building and Environment, Energy and Buildings, Urban Climate, and Computers, Environment and Urban Systems, accumulating 892 citations · h-index 15 · i10-index 17 (Google Scholar). I hold a software invention disclosure (SpaceBrain, Ref. 2021-140) and have contributed to a SGD 1,686,100 competitive grant proposal as Co-PI. The vast majority of these citations have accrued since 2021, reflecting accelerating impact, and I am on track to exceed h-index 20 within three years.

See the Publications page for the full record.