Rahul Marchand
I'm a final-year engineering student at the University of Oxford, doing my master's project in Maike Osborne's Bayesian Governance Lab. I study the internal structures behind goal-directed behaviour in AI agents. This builds on some exploratory interpretability work I carried out over the summer at the Witt Lab. I'm also co-mentoring a SPAR stream on the science of evaluations.
Before that I built a benchmark of container sandbox escapes for LLM agents with the UK AI Security Institute, and worked on machine learning for tuning quantum dot devices in Natalia Ares's group.
Outside work I coxed for Oxford against Cambridge in the 2024 and 2025 lightweight Boat Races. I'm Belgian and Sri Lankan, and grew up in London.
Publications
- QArray+: A Physics-Informed GPU-Accelerated Simulator for Quantum Dot ArraysP. Vaidhyanathan, B. van Straaten, A. Petrillo, R. Marchand, E. De Nicolo, M. Veldhorst, B. Khailany, T. L. Patti, N. AresarXiv, 2026
- Action-Factored Multi-Agent Reinforcement Learning for Scalable Quantum Device TuningE. De Nicolo*, R. Marchand*, C. Carlsson, P. Vaidhyanathan, N. AresarXiv, 2026
- Quantifying Frontier LLM Capabilities for Container Sandbox EscapeR. Marchand, A. O Cathain, J. Wynne, P. M. Giavridis, S. Jennings, F. Tuxworth, T. H. Dur, S. Deverett, J. Wilkinson, J. Gwartz, H. CoppockICML 2026, oral ยท code
- End-to-End Analysis of Charge Stability Diagrams with TransformersR. Marchand*, L. Schorling*, C. Carlsson, J. Schuff, B. van Straaten, T. L. Patti, F. Fedele, J. Ziegler, P. Girdhar, P. Vaidhyanathan, N. AresarXiv, 2025
* equal contribution
