I enjoy pushing the boundaries of AI systems while working to understand their behaviour and harness their potential safely to expand what’s possible.

I’m currently post-training computer use agents and designing their training and evaluation systems at Salesforce, which I joined through its acquisition of Convergence.

Previously, as Unitary’s first employee, I led multimodal model research for harmful content detection and created Detoxify. Before that, I researched adversarial robustness at the University of Oxford and developed computer vision systems for detecting sight-threatening diseases at Visulytix.