Zijian Wang

Dr. Zijian Wang is an independent postdoctoral researcher at the TUM Georg Nemetschek Institute (TUM GNI), Technical University of Munich, advancing construction digitalization through AI and graph-based approaches to Building Information Modeling (BIM).

His educational background includes a PhD from Technion – Israel Institute of Technology (Marie Curie Early-Stage Researcher in EU Horizon 2020 Cloud BIM project), a master's in Computer and Machine Vision from Cranfield University, a master's in Civil Engineering from Central South University, and a bachelor's in Civil Engineering from Chongqing Jiaotong University. He has also conducted research at the University of Cambridge and industry work at Trimble Finland.

Dr. Wang's research centers on Building Information Graphs (BIGs), transforming design data into graph structures to enable machine learning applications. His work addresses critical gaps in BIM interoperability through semantic enrichment using graph neural networks, with applications in construction safety monitoring, multidisciplinary change propagation, and automated consistency maintenance. Specific projects include PPE detection via deep learning, tunnel temperature data mining, and graph-based object classification.

His publication trajectory (2021-2025) demonstrates increasing sophistication in graph-based BIM solutions, evolving from foundational room classification to complex cross-domain interoperability frameworks and generative AI applications for construction planning constraints.

He has received significant recognition:

• Thorpe Medal, European Council on Computing in Construction (EC3) (2023)
• Eastman Best Ph.D. Paper, Int Joint Conference CIB W78 - LDAC (2021)

Dr. Wang supervises theses in TUM's Software Lab and contributes to EU-funded research initiatives. His current BIGs research aims to establish graph-based foundations for generative design tools, fostering more collaborative and intelligent construction workflows through explicit relationship modeling and multi-modal data embedding.