About

Welcome! This is the house of jgcarrasco on the Internet. It contains a compilation of writings, posts and other information that I find interesting.

I’m a recent PhD in Computer Science, where I mainly worked on the mechanistic interpretability of Large Language Models (LLMs), focusing on its applications (e.g. vulnerability detection, pruning). I have a Physics background but became interested in Deep Learning and decided to pursue a PhD, and it turned out to be an amazing oportunity to learn about a lot of things.

Currently, my main topics of interest are:

• Using LLM agents for practical applications. Current LLMs are powerful enough, in some cases we just need proper scaffolding/harnesses.
• GPU Programming.
• Linux, self-hosting, embedded systems. Deploying LLMs locally.