About
Hey! I'm Matthew. I'm a physicist by training, now focusing on getting into AI safety. Other interests of mine include tennis and running. This website serves as a platform where I share my thoughts, research, and insights on topics that interest me.
CV
View CV (PDF)Background
I have a background in accelerator physics, with particular interest in beam dynamics. Currently, I'm working as a physicist at UCLH (London, UK) where I engage in research projects that aim to improve the efficiency and quality of care for patients treated with radiotherapy.
Research Interests
My primary research interests include:
- AI Safety, including mech interp, alignment, control and monitoring
- Proton Beam Angle Optimisation in Proton Radiotherapy
- Accelerator physics, including RF and particle beam dynamics
- Statistical modelling e.g. Stochastic Differential Equations to simulate particle transport
Publications
- Patil, Shiv & Karabacak, Mert & Southerby, Matthew & Margetis, Konstantinos. (2026). Embracing Large Language Models for Medical Applications, Part II: Building a Framework for Clinical Stewardship.. Cureus. 10.7759/cureus.115367. ResearchGate
- Southerby, M. et al. (2025). Fast SDE-based Monte Carlo dose calculation for proton therapy validated against Geant4. arXiv:2511.03115. arXiv
- Southerby, M., & Apsimon, R. (2024). Beam Dynamics Framework Incorporating Acceleration to Define the Minimum Aperture in Two Focusing Schemes for Proton Radiotherapy Linac. arXiv:2403.10212. arXiv
- Southerby, M. (2023). Beam Dynamics Studies and RF Cavity Design for Proton Radiotherapy Machines. PhD Thesis, Lancaster University.
- Southerby, M. et al. (2022). Multiobjective optimization and Pareto front visualization techniques applied to normal conducting rf accelerating structures. Physical Review Accelerators and Beams, 25(6), 062002. DOI