UChicago CS PhD
More capable AI is not the same thing as safer AI. Safety starts with understanding. I study how AI agents interact and how to make it safe.
“AI is advancing faster than we can control, and the people building it do not truly understand it.”
— Geoffrey Hinton, 2024 Nobel Prize in Physics
Research, infrastructure, and product engineering.
More capable AI is not the same thing as more trustworthy AI.
My interest in AI safety grew out of medical research, where a claim does not mean much unless it can be examined, challenged, and tested. What kind of knowledge are we producing if we cannot examine how it was made? That is part of why I care about evaluation, verification, and interpretability for agentic systems.
I care about systems whose reasoning can be followed closely enough to be questioned. Can we see what evidence shaped a conclusion? Can we identify where the reasoning started to weaken? Can we tell when a polished answer is covering over a fragile process? A system can sound coherent and still fail in ways that are hard to see at first.
I do not think rigor can be separated from human taste and the willingness to keep asking questions. Good judgment often begins when something sounds complete, but does not feel settled. What is missing. What has been assumed. What has not been tested. In science and medicine, that habit of questioning is part of care.
I believe science must serve humanity's welfare rather than corporate or individual greed. I believe that the role of an inventor is to perform a sacred duty for the collective elevation of human civilization and the collective good. If science does not serve the collective good, what is progress for?
I want to build things that make AI safer to use in science and medicine, help create a research environment where speed never comes at the cost of care, and contribute to scientific advancement in ways that serve the collective good.
Reach me through research, product, or founder work.
