starting point: 

How can Machine Learning–enabled, data-driven design frameworks enhance collaboration, data organisation, and content sharing in the Architecture, Engineering, and Construction (AEC) sector, particularly within performance-based design workflows?

The Architecture, Engineering, and Construction (AEC) sector is responsible for nearly 40% of global CO2 emissions, about 36% of global final energy use, and more than half of global electricity demand, according to the IEA’s Global Status Report for Buildings and Construction (“IEA, 2019 Global Status Report for Buildings and Construction,” 2019). Alongside these pressures, the sector consumes vast natural resources and generates significant landfill waste (Pan et al., 2023). Early-stage design decisions are critical to reducing these impacts, yet current workflows remain fragmented, heavily manual, and reliant on time-intensive simulations. Currently, less than 20% of firms actively use Artificial Intelligence (AI) or Machine Learning (ML) tools tailored for this sector (Lystbæk, 2025), even though these technologies have the potential to transform practice by analysing complex datasets, predicting outcomes, and guiding sustainable decisions when they are most cost-effective. Barriers such as siloed data systems, poor interoperability, and fragmented communication undermine collaborative practice. At the same time, the sector faces growing pressure to deliver buildings that meet ambitious operational and embodied carbon targets. These challenges require a fundamental rethinking of how design data, performance metrics, and collaborative processes are interlinked from the earliest stages of design.

research focus:

This project explores how ML-enabled, data-driven frameworks can strengthen collaboration within performance-based design workflows. It focuses on the integration of real-time performance evaluation into design processes, while also advancing a data-centric architecture in which structured, shareable, and interoperable datasets can support the entire project lifecycle. By combining these strands, the research positions ML not as a supplementary tool but as an integral driver of adaptive, collaborative, and multi-objective design practice. 

project summary: 

The research develops a multi-scalar optimisation framework that embeds ML directly into performance-based design. This framework operates across the micro (material and component), meso (building), and macro (urban) scales, creating feedback loops in which insights at one level inform decisions at another. By automating the organisation of geometric, material, and performance data, and ensuring interoperability with BIM and parametric tools, the framework enables design teams to pursue multiple objectives simultaneously, from energy efficiency and carbon reduction to cost control and constructability. Collaboration is central; ML-enabled systems are designed to gather inputs from diverse stakeholders and provide predictive insights early enough to shape foundational design choices. In practice, this reduces redundant manual coordination, aligns participants within a unified data environment, and allows performance scenarios to be tested long before costly commitments are made. 

methods and validation 

The research follows a Research through Design methodology, combining structured literature reviews with computational prototyping, surrogate modelling, and multi-scalar optimisation. Hackathons and industry case studies provide practical testing grounds, while methods such as Physics-Informed Machine Learning (PIML) (Ma et al., 2025), Graph Neural Networks (GNN), and Reinforcement Learning (RL) expand the framework’s adaptability. Together, these approaches demonstrate how ML can be meaningfully embedded into early-stage workflows, bridging the persistent gap between design intent and realised performance. 

impact 

The project aims to deliver both theoretical and practical contributions. It will clarify the role of ML within design optimisation, while also producing open-source tools, tested workflows, and guidelines for industry adoption. By improving collaboration, data organisation, and performance-based decision-making, the research supports the sector’s digital transformation goals and accelerates progress toward a low-carbon, high-productivity future (Kookalani et al., 2024). 

references 

“IEA, 2019 Global Status Report for Buildings and Construction” (2019). 

Kookalani, S. et al. (2024) “Trajectory of building and structural design automation from generative design towards the integration of deep generative models and optimization: A review,” Journal of Building Engineering, 97, p. 110972. Available at: https://doi.org/10.1016/j.jobe.2024.110972. 

Lystbæk, M.S. (2025) “Machine learning-driven processes in architectural building design,” Automation in Construction, 178, p. 106379. Available at: https://doi.org/10.1016/j.autcon.2025.106379. 

Ma, Z. et al. (2025) “A review of physics-informed machine learning for building energy modeling,” Applied Energy, 381, p. 125169. Available at: https://doi.org/10.1016/j.apenergy.2024.125169. 

Pan, Y. et al. (2023) “Building energy simulation and its application for building performance optimization: A review of methods, tools, and case studies,” Advances in Applied Energy, 10, p. 100135. Available at: https://doi.org/10.1016/j.adapen.2023.100135.