About

Hi, I’m Kerem! I’m a Research Assistant at Northeastern University, working in the Bau Lab. I completed my master’s in CS at Northeastern, supported by the Fulbright Program.

My research focuses on mechanistic interpretability and training dynamics. I’m broadly interested in how mechanisms and representations in language models emerge during training, and how manipulating training procedures can reveal which properties of the data and training regime drive the learning of these mechanisms. Recently, I’ve been thinking about applications of scalable interpretability to self-explanations.

Before Northeastern, I completed my bachelor’s degree in CS at Bilkent University.