Research


We build AI and machine-learning methods and apply them to high-impact problems in biomedicine and engineering. Our work spans the full stack — from foundation and generative models, through scientific machine learning, to trustworthy, physics-aware AI — grounded in deep domain expertise in mechanics and simulation.


AI in Biomedicine

We develop machine-learning and in-silico models that bridge engineering and clinical medicine — turning medical images and patient data into actionable, decision-support tools. Our work spans cardiovascular disease, ophthalmology, and soft-tissue mechanics, combining learned models with physics-based simulation for reliable, patient-specific predictions.

Key topics:

  • Machine learning for medical imaging (retinal fundus, cardiovascular)
  • Patient-specific, in-silico models for risk stratification
  • Virtual patient populations and digital twins
  • Physics-informed models of soft tissue and haemodynamics

Application highlights: Cardiovascular AI Retinal imaging In-silico medicine

Scientific Machine Learning

We design neural architectures that learn the physics of engineering systems — graph neural operators, physics-informed neural networks, and surrogate models that replace expensive simulations while respecting the underlying mechanics. These methods make high-fidelity analysis fast enough for design, optimisation, and uncertainty quantification at scale.

Key topics:

  • Graph neural operators for PDEs and mesh-based systems
  • Physics-informed and physics-constrained learning
  • Surrogate modelling for multi-physics simulation
  • Uncertainty quantification and robustness under distribution shift

Application highlights: Neural PDE solvers Wave-structure interaction Digital twins

Foundational AI

We explore how modern AI — large foundation models, generative models, and agentic systems — can transform how engineering and scientific work gets done. From generative design and inverse problems to AI agents that drive simulation and analysis pipelines, we bring contemporary CS-style AI to bear on engineering practice.

Key topics:

  • Foundation models for scientific and engineering data
  • Generative models for design and inverse problems
  • Agentic AI for simulation, analysis, and engineering workflows
  • LLM-assisted modelling and scientific tooling

Application highlights: Generative design AI agents Scientific tooling


Former & Foundational Areas

Our AI methods are grounded in a strong foundation of computational mechanics. These earlier research directions remain part of the lab’s heritage — the publications and software stay fully available and continue to inform our scientific machine-learning work today.

Classical finite element methods Space & deployable structures Wave-structure interaction Soft-tissue & contact mechanics

See Former Areas for details.