Focused primarily on AI model evals and AI governance. I am currently starting a new for-profit, safety-focused AI evals org, Suav Tech. Our initial focus is to build additional general-purpose scaffolding to run evaluations on UK AISI's Inspect framework and create new benchmarks for long-term planning, strategic reasoning, and novel knowledge creation. We are currently fundraising.

In addition, I am currently a member of the working group on international AI governance at the Alva Myrdal Centre for Nuclear Disarmament at Uppsala University, working on publications related to the influence of nuclear governance upon AI governance and the impact of AI on nuclear verification. I also work as a research engineer at Equistamp, where I have been helping UK AISI build out ControlArena and worked on other client projects.
Previously I have contracted with METR for baselining, model interaction and data annotation, facilitated courses on AI Governance and Intro to TAI by BlueDot Impact, and worked on projects mapping the AI governance landscape of India, and studying "Atoms for Peace" and how AI governance can learn from it.