Understanding language models.
Building safe, reliable AI.

I deeply admire explorers such as Jacques Piccard and Jill Heinerth—their curiosity, courage, and willingness to venture into the unknown. That spirit resonates with how I approach language models. For me, diving into the dark means exploring their inner workings: looking beneath their outputs, questioning familiar explanations, and seeking understanding grounded in evidence, while remaining open to what we have yet to discover.

For me, exploration also carries responsibility. I want that understanding to become knowledge others can build on, and to help make AI systems safer and more reliable. I care about how these systems affect people, whose needs they serve, and who bears their risks. My hope is to contribute to AI whose benefits are shared more widely and fairly.