
Selected Projects
Sustainability
Urban Building Energy Modeling
-
Developed Urban Building Energy Models (UBEMs) for cities around the world to support policy interventions to meet their carbon emissions reduction targets.
-
Geometry: GIS, Rhino3D, Python
Simulation Engine: EnergyPlus
Simulation Interface: UBEM.io, UMI, ClimateStudio, Grasshopper
Data Science / Machine Learning: Python
Urban Analytics - Boston TreeScape
-
Combined various public datasets to analyze outdoor thermal comfort and walkability in Boston.
-
Data Science / Machine Learning: Python
Geospatial: Python, Mapbox, Carto, Leaflet
Geometry: Rhino3D
Design - Daylight Optimization
-
Developed computational generative design workflow and scripts to optimize daylight (and shading/reflective devices) for buildings.
-
Rhino3D, Grasshopper, C#
Interventions for Heat Risk in Informal Settlements
-
Experimentation, field measurements, and physics-based modeling to develop cost-effective interventions for underprivileged communities in informal settlements across the Global South.
Country-Level Solar Energy Potential Study
-
Integrated multiple public datasets to comprehensively assess solar energy potential for both rooftops and facades across diverse residential areas in Singapore. The methodology incorporated physics-based simulations and employed a Generative Adversarial Network (GAN) model to expedite result generation and analysis
Led by Yifei
-
Data Science / Machine Learning: Python
Geospatial: QGIS, Mapbox
Geometry: Rhino3DSimulations: ClimateStudio, Ladybug/Honeybee
Technology (AI + Others)
Impact of Visual Elements on Thermal Perception
-
Leveraged computer vision models, including ResNet and Vision Transformers, in conjunction with custom-developed dissimilarity metrics and targeted survey methodologies. This approach was used to analyze and quantify the influence of various visual elements on human perception of outdoor thermal comfort
Led by Lujia
-
Data Science / Machine Learning: Python
Crowdsourced Campus Digital Twin
-
Crowdsourced campus digital twin for research and teaching.
Led by Lester, Jolyn
-
Geospatial: Mapbox, CesiumJ
Geometry: Rhino, OSM, RevitFrontend: React
Data Science / Machine Learning: Python
Simulations: ClimateStudio
Deep Tech Hardware - Novel Renewable Energy Device
-
Advanced clean energy system that harnesses solar energy to generate electricity with up to 3 times higher efficiency compared to conventional solar panels. This novel technology also boasts a significantly reduced carbon footprint throughout its lifecycle.
Led by Sparsh
Computer Vision - Urban Features Extraction
-
Using computer vision techniques to extract features of the urban built environments through datasets such as satellite imagery, for multiple purposes and use cases.
-
Models: Python/Tensorflow
Geometry: Rhino3D, Grasshopper
Satellite and other data sources confidential.
Urban Analytics - Boston Crime Rate
-
Combined various public datasets to analyze crime rates in Boston, as well as the key factors affecting crime in various neighborhoods. A machine learning prediction model was developed to predict crime rate in different regions.
-
Data Science / Machine Learning: Python
Geospatial: Mapbox, Carto, Leaflet
Generative Design - Parametric City
-
Developed a computational generative design script to automatically explore planning variations and simultaneous optimizing on-site solar energy generation (rooftop PV).
-
Rhino3D, Grasshopper, C#
Building Information Modeling
-
Computational Building Information Modeling (BIM) throughout the building delivery process, including design optimization, cost mapping (automated cost plans), clash detection and prioritization, and value engineering
-
Models: Revit, CostX, NavisWorks
Code: Python, C#, Excel VBA
Consulting & Technology Roadmapping
-
Consulting project for a smart infrastructure MNC to map the global quantum computing landscape, including key stakeholders, market trends, patent/intellectual property, as well as technology roadmap development.
Policy
Policy - Predicting Startup Funding & Success
-
Combined various public datasets to build data-driven machine learning models that predict startup funding and success in various cities. Findings support policy interventions and levers that spur startup and innovation growth.
-
Data Science / Machine Learning: Python
Statistics: R
National Robotics & Manufacturing Policy
-
Designed and developed national policies for robotics and the advanced manufacturing sector, including interventions and levers to enhance R&D, commercialize technology, grow high-tech startups, and attract international corporations.