Ayush

Notes on systems, code, AI and Inference

I take apart complex systems and write up how they actually work — the architecture, the trade-offs, and the parts the README leaves out.

About me Email

Writing

inference deep dive

DeepInfer Explained — a complete field guide to DeepSeek's inference engine

Everything in the DeepInfer repository, explained from zero: the architecture, a map of all 342 files, the request lifecycle, and every concept it drops — FlashMLA, DeepGEMM, DeepEP, DSpark, SWA bounded replay, EPLB and the rest.

Read →