I am a researcher at METR (formerly ARC Evals) working on evaluations of the capabilities of language models. You can learn more about our work from our recent report.
Previously, I was a research resident at Redwood Research, where I worked on Mechanistic Interpretability of Language Models, and before that, I was a Master's student in Computer Science at Stanford, where I split my time between Machine Learning (researching Natural Language Understanding for virtual assistants at the Stanford Open Virtual Assistant Lab), and Human Learning (researching mathematical cognition with human subjects at the Parallel Distributed Processing Lab).
I was also a Teaching Assistant for Stanford's Computer Science department. Courses I have been a Teaching Assistant for include CS 379C (Computational Models of the Neocortex), CS 107 (Computer Organization and Systems), and CS 103 (Mathematical Foundations of Computing).
I spent my undergrad years in the Mathematics and Philosophy departments at Stanford. Back then, I was mostly interested in Logic, Topology, and Philosophy of Science. These days, I tend to be a bit more practical, but I still have a soft spot for those subjects.
My github is @satojk, and you can reach me at satojk at cs.stanford.edu.