Interests
Subjects I keep coming back to.
Bayesian workflows
Priors as hypotheses, posteriors as arguments. I'm interested in pipelines that make uncertainty a first-class output, not an afterthought.
Causal inference in business
Most decisions inside companies are causal questions in disguise. DAGs, difference-in-differences, synthetic controls, whatever the setting demands.
AI safety
How do we keep increasingly capable systems reliably doing what we actually want? I'm drawn to the measurement and evaluation side: telling whether a model has internalized a rule versus pattern-matched around it.
Writing tools for thinking
Plain text, local files, a few well-aimed scripts. Less magic, more leverage.