Advising research labs

AI accelerates discovery. Science still depends on principles that make its operation trustworthy.

---

AI can compress years of scientific work into months, generating hypotheses, running simulations, and analyzing data at a scale that humans in a lab cannot manage manually. That speed creates risk. Labs that treat AI as a shortcut around rigor, reproducibility, and independent verification produce results that do not hold up, however fast they arrived.

I work with research labs, universities, and funders to build frameworks for using AI as a scientific tool rather than a substitute for scientific judgment. That means helping teams decide where AI-generated hypotheses need independent verification, where reproducibility standards have to change for AI-assisted methods and where they cannot, and how to keep a human researcher accountable for conclusions a model helped produce.

Why me

I am an Ethics and Technology Practitioner Fellow at Stanford's McCoy Family Center for Ethics in Society, where my current work develops frameworks for AI's role in scientific discovery. Before Stanford, I spent five years as a research scientist and tech ethics advisor inside Google's AI teams, and I hold a DPhil from Oxford's Internet Institute in engineering and research ethics. That combination positions me close enough to the technical process to know what a model can and cannot be trusted to do, and trained enough in research integrity to say why the distinction changes what a lab should publish.

Who this is for

This is for labs integrating AI into their research pipeline who need to know where to hold the line, and for universities and funders setting policy for AI use across many research groups at once. It is also for any team that has noticed a gap between how fast its AI tools let it move and how confident it still is in what those tools produce.