About
I work to help AI go well — building the science of how we measure and understand what advanced AI systems can actually do, backed by a research career designing quantum algorithms for some of chemistry's hardest problems.
AI evaluation & safety. I help build rigorous, predictive methods for assessing what AI systems can do. I'm a co-author of General Scales Unlock AI Evaluation with Explanatory and Predictive Power (Nature, 2026), which introduces general scales — built on 18 rubrics of cognitive and intellectual demands — that explain what AI benchmarks really measure, profile the strengths and limits of language models, and predict performance on new and out-of-distribution tasks. As Director of the International Programme on AI Evaluation I lead international efforts to assess AI capabilities and safety, and at the EU AI Office I work on frontier AI safety, where technical evaluation meets policy.
Quantum algorithms. I design fault-tolerant algorithms for simulating chemistry and materials on quantum computers. My most recent work, Theory and Practice of Trotter Product Formulas for Quantum Chemistry (2026), develops the most efficient product formulas for electronic Hamiltonians in second quantization, cutting the real cost of Hamiltonian simulation. I also created GradDFT, a JAX library for the differentiable design of exchange–correlation functionals, with earlier work on protein folding (QFold), linear-programming optimization, and lithium-ion battery simulation.
Experience & education
Research
Projects & interactive essays
Writing
Contact
Have a question, or want to get in touch about AI evaluation, AI safety, or quantum computing? Send a message — it comes straight to my inbox.