Hi, I'm Ayush. I like understanding how big systems really work, and the fastest way I
know is to read the source. This blog is where I write down what I find, mostly about
LLM inference and the engineering that makes models fast and cheap to serve.
What I write about
- Reading real codebases to see how large systems work under the hood
- Inference and performance: serving, scheduling, KV caches, and making models cheaper to run
- Notes on papers and the problems I had to work out the hard way
Elsewhere