xianyunshi
An animated scene: a white-haired elf watches clouds drift past, rapeseed flowers in the foreground.
闲云 · idle clouds

Notes on random ideals, deep work, measurement , inspiring lessons during ML/LLM, diffusion project journeys. Almost like deep algorithmic work and system design work which i sure would forget in a few months. Will want to write other casual things too, soon.

elsewhere — github , portfolio , linkedin

streaming notes: async def, threadpools, and bridging a synchronous pipeline into SSE

a question answering service sat on a blank spinner for a minute or more per query, and the fix went through four changes that were all really about one seam between synchronous work and an async framework

reranker evaluation notes: cross-encoder vs BM25 on a financial-filing RAG pipeline

a cross-encoder and BM25 both went into a real financial-filing RAG system, and the project's own investigation found the reranker made results worse, for reasons that turned out to be about pruning, not the model

embedding model notes: comparing providers and catching non-determinism without a baseline

five times I measured something that refused to hold still, and the only thing that saved me was measuring something else beside it

container cost notes: splitting cold start, import, and cache costs out of one measurement

three times a single reading hid the cost that actually mattered, and each time the fix was to measure again from somewhere else

multi-turn chat notes: why a stateless API has to resend the whole conversation every turn

multi-turn feels native to build because it is list.append. every genuinely hard part of it lives somewhere else.

polars learning notes: lazy scanning, streaming collect, and a 39GB read that crashed

thirteen things that actually stuck from three practice notebooks, and the eager read that undid all of them at the one file where it mattered

Kim Dokja, arms raised, declaring that he is no longer mentally ill.