Research Scientist at Apollo Research (June 2025–Present, Zurich/London): Focused on evaluating AI scheming propensities and AI control.
AI Safety Researcher (Jan 2025–May 2025, Remote): Researched AI sandbagging and control, assessing monitors’ ability to detect sandbagging and sabotage.
Resident at Mantic (Oct 2024–Present, London/Remote): Contributed to building an AI superforecaster.
Independent Research on AI Sandbagging (Aug 2024–Oct 2024, London/Remote): Continued sandbagging research with a grant from the AI Safety Fund, collaborating with Francis Rhys Ward and Felix Hofstätter.
Research Scholar at MATS (Jan 2024–Jul 2024, Berkeley/London/Remote): Worked on strategic underperformance (sandbagging) in general-purpose AI under mentorship of Francis Rhys Ward.
Co-founder and Co-director at ENAIS (Dec 2022–Present, Remote): Building European coordination in AI safety.
SPAR Participant (Feb 2023–May 2023, Remote): Focused on evaluating the shutdown problem in language models at UC Berkeley’s Supervised Program for Alignment Research.