Hi, I am Karthik, an AI alignment researcher who approaches problems as both a computer scientist and a theoretical physicist. This mix has let me move quickly and adapt across very different worlds: competitive programming, data science in industry, a physics PhD, and now LLM interpretability and AI safety. Over the last few years my work has centred on one question: what is a model actually doing internally, and how do we use that to make it more aligned with human values and preferences?