Digital twins in scientific domains are traditionally built around physics-based simulations calibrated with measurements. As sensing technologies and data volumes grow, these systems require computational methods capable of inferring structured representations of system state from heterogeneous observations in ways that interface naturally with simulation workflows.

In this talk, Dr Averkiou will discuss how structure-aware artificial intelligence can contribute to such systems by learning explicit geometric and relational representations from complex data. These representations support structured inference, semantic modeling, and learning-based components that can operate alongside simulation models.

Dr Averkiou will illustrate this through examples in the built environment, where learning-based methods infer structured geometric and semantic representations from imagery and 3D data. He will also briefly discuss structure-conditioned generative models as examples of incorporating explicit geometric priors within data-informed digital twin pipelines.

Dr Averkiou will conclude by discussing future directions in which structure-aware AI can complement existing strengths in simulation and high-performance computing at CaSToRC, through learning-based components that interface with scientific modeling.

Dr. Melinos Averkiou is an Associate Research Professor and Group Leader of the Visual Computing Group at the CYENS Centre of Excellence in Cyprus. He received his PhD in Computer Science from University College London in 2015 and has held research positions at Stanford University, SIAT China, and the University of Cyprus.

His research focuses on structure-aware artificial intelligence and neural methods for modeling and understanding complex 3D data, with applications spanning computer vision, remote sensing, and large-scale geometric modeling. He also oversees Prometheus, CYENS’ peta-scale AI computing infrastructure.