Antonietta Ambuehl

I hold an MMath and an MSc by Research from the University of Oxford, where I developed and published methods for simplifying complex mathematical models. While working as a Cost Estimator in the UK Civil Service, I completed an MSc in Theoretical Physics at King’s College London, focusing on quantum field theory and topological entanglement entropy. In my professional role, I applied mathematical modelling and machine learning to budget forecasting and risk analysis.

Large amounts of energy are released when light atomic nuclei fuse to heavier ones. For fusion to happen, the fuel is heated to extremely high temperatures. This state is called a plasma, where electrons become separated from the nuclei so that these can collide and fuse. In fusion plasmas, high‑energy electrons can carry heat over long distances, a phenomenon known as non‑local thermal transport. To improve the design, interpretation, and efficiency of fusion experiments we need to be able to predict how energy moves through the plasma, how hot it stays, and how it interacts. However, experimental measurements of non‑local transport are limited and mathematical models, simulations of the experiment, are complex and computationally expensive.

Physics‑Informed Gaussian Process Regression (PI‑GPR) is a modelling approach that can make accurate predictions even with limited simulation or experimental data. In my project I will develop a mathematically rigorous formulation of PI‑GPR that captures the non‑local thermal transport effects and provides quantified uncertainty of the predictions. This goes beyond applying existing machine‑learning models, focusing instead on building the mathematical structure needed for physically consistent modelling. I aim to contribute computational tools that accelerate fusion research by balancing speed and accuracy. The methods developed could be useful in other engineering, physics, and computational modelling domains.

Supervisors

• Dr Jiannan Yang