About Kunvar

Hello, welcome to my corner of the internet. I’m Kunvar. I am a machine learning engineer. I specialize in taking apart neural networks and figuring how what they learn.

This general line of work is often called “Interpretable AI”, or “explainable AI” or “XAI” in literature. I am presently working on a type of interpretability where I rigrously reverse engineer AI systems, specifically transformer-based models, to understand the structure of their knowledge. This is also called “Mechanistic Interpretability”.

You can find more information about my work in mechanistic interpretability on my website mechinterp.

You can find links to my socials here.