I’m Andrea, a software developer working on local LLM inference.

I build Hebrus, a Metal-first inference engine for Apple Silicon with adaptive SSD streaming. Most of the work starts with a model that does not fit, a machine that should not be enough, and a measurement I do not trust until I can reproduce it.

Before this I worked on cloud infrastructure. Some habits survived: make the bottleneck visible, test the obvious answer, and keep the claim narrower than the evidence.

I also work on forgequant, benchy, and an experimental ds4 fork. Model artifacts and experimental builds live on Hugging Face. This site holds the field notes: what ran, what stalled, and which promising idea got slower when measured properly.

Away from the terminal, I mix techno, UK garage, and whatever other underground neurospicy shit has made it into the folder that week.

Email: hey@andreabor.io.