My research is largely focused on mechanistic interpretability and reverse engineering intelligent (learning) systems—both artificial and biological—to understand how complex behavior and cognition emerge at a mechanistic level. This research spans multiple disciplines (neuroscience, machine learning, and healthcare) and levels of organization, from circuits to emergent properties and higher-level theories of computation. My long-term goal is to apply these insights to advance AI safety, drug discovery, and treatments for neurological disorders.

Alongside theoretical work, I am also interested in developing efficient software tools to help study large datasets and models at scale. I am proficient in Python and most of its scientific computing libraries, though am interested to learn more about distributed computing, writing custom CUDA kernels, and high-performance machine learning.

When I'm not writing or doing research, you can likely find me training at Minneapolis Bouldering Project.