I am a Research Fellow at Imperial College London, where I work on safe
My work primarly concerns the development of formal methods for assessing the
safety and robustness of AI systems. It covers both symbolic
and connectionist systems, including
- optimisation-based approaches for the verification of adversarial
robustness properties of neural networks,
- logic-based approaches for the verification of the temporal-epistemic properties of multi-agent systems, and
- hybrid approaches for the verification of strategic properties of neuro-symbolic systems.
Part of my research implements several core components of the verification
engine of Safe Intelligence, where I partly work as a Senior Research Engineer.