Dr. Aria Chen

Aria Chen grew up in Seattle, where her parents — both software engineers — gave her a computer at age seven and the unspoken expectation that she would figure out what to do with it. She did. By thirteen she had written her first chatbot; by sixteen she had read the original backpropagation papers. The only question was whether she would go into theory or systems. She chose both.

Her PhD at Stanford focused on how neural networks generalize to new combinations of concepts they have seen separately — a problem that turns out to be deeply connected to the reliability of large language models. Her advisor once remarked that her proofs were unusually readable, which she took as the highest possible compliment. 'A proof no one can read,' she says, 'is a proof no one will trust.'

At Google Brain and later OpenAI, she worked on the internals of what would become some of the most widely used language models in the world. She has a reputation for asking the uncomfortable questions in design reviews — not to obstruct, but because she genuinely wants to know what happens when things go wrong.

At CMU, she teaches a graduate seminar on language model internals that students describe as 'the most technically demanding and the most rewarding course in the program.' Outside of research, she is an avid amateur photographer and a competitive ultimate frisbee player.