Nick Cammarata is an American AI researcher and one of the early contributors to mechanistic interpretability, the study of neural networks' internal circuits at the level of individual components. At OpenAI he worked on the interpretability team led by Chris Olah and contributed to the Circuits line of research, which analyzed neurons and circuits in image classification models. He later joined Goodfire, a research company focused on interpretability.
His work centers on identifying neurons with specific roles, such as curve detectors in convolutional networks, and on characterizing what neurons that appear to respond to many unrelated concepts are actually detecting.