About

I’m Abhi. Right now, most of my time goes toward AI research, specifically mechanistic interpretability. Language models keep getting more capable, and we understand less and less about how they actually work. That gap is what I’m working on in whatever ways I find interesting.

I’m particularly interested in interpretability methods that scale cleanly with compute. I’m currently doing research on introspection in LLMs at UVA.

Before all of this, I spent all my time obsessively doing CTF competitions, mostly binary exploitation and reverse engineering. Some of those writeups are still on here. I still find what I learned during that time incredibly useful, but my interests have just shifted.

Tech Stack

This site is built with Astro, TypeScript, TailwindCSS, and Markdown. Hosted on Cloudflare Pages. Based on astro-paper.