February 18th, 2026, Wednesday, 3PM. Room TG23.
Speaker: Elisabeth Fink.
Title: AI for Science: From Accelerating Materials Discovery to Stress-Testing Post-Quantum Cryptography
Abstract: Applications of AI have become widespread and the UK has defined “AI for Science” as a national strategy. Within this, the UK aims at becoming a leader specifically in scientific applications of AI, including several high impact topics. In this talk I will outline the current state of the field in some selected areas. Then I will present some of my own research.
The first example will be for AI in material science. HPC simulations like the discrete element method are common tools in engineering to understand material flow. They are however computationally very expensive. AI tools have been developed in the past few years to approximate their extensive calculations. While there are several examples of such predictive models, they are usually constraint by the maximal number of particles and sometimes require large amounts of data (and hence energy) for training. The presented solution will be a lightweight framework for tracking particle flow in a grid-based approach.
In the second part of this talk I will present an application of AI to mathematics, which studies systems that can be used for post-quantum cryptography. Classical cryptographic systems rely on the computational complexity of prime factorisation or the discrete logarithm. Quantum computers leverage algorithms which allow for these problems to be solved in polynomial time. To address this, more complex mathematical protocols using infinite groups have been proposed. While certain problems in these groups are believed to be difficult, there are some first indications that AI can be used to gain a computational advantage and hence expose vulnerabilities. In this talk I will give an overview of current developments and give very recent research results. This will highlight how AI can be used in mathematics to advance the understanding of infinite groups and hence expose possible vulnerabilities in post-quantum cryptographic systems.
Bio: Elisabeth Fink holds a DPhil in Pure Mathematics from the University of Oxford. Before this, she graduated with distinction in Mathematics and Computer Science from the University of Technology in Vienna (TU Wien). Her experience includes several years as a Postdoc at the ENS Paris as well as the University of Ottawa. Dr Fink gained valuable industrial insight when working as a Data Scientist in logistics, banking and a research centre for pharmaceutical engineering. Since 2025, she has been a Lecturer in Computer Science at Middlesex University London.