Zhiyuan Lin

PhD candidate / ML & Systems

A word from my assistant

RESEARCH INDEX

Papers, tools, and experiments share one index. Every entry keeps a reproducible entry point.


My desk: half papers, half GPU fan noise
My desk: half papers, half GPU fan noise

I'm Zhiyuan Lin, a PhD candidate working on machine learning systems — I care both about what models can do and how efficiently they run.

Beyond research, I maintain a few small open-source tools, follow the frontier papers, and believe in two things:

  1. Reproducibility matters more than leaderboards;
  2. The best system is one you can read through in an afternoon.

Curious about the tools and methods behind my writing and research? See the feature tour — this site itself is the demo.

VISUAL LAB

Tiles support fast scanning; archive cards preserve retired experiments.

ArchivedOld experiment AThe service stopped, but the conclusions and config remain useful.
ArchivedTravel Log 2025Old material still belongs in a research index.

3,677 GitHub contributions in the last year

stlin256/OpenHomepage-V2Lightweight, magazine-style static personal homepage generatorastrobilingualgithub-pagesmarkdownpersonal-homepageportfolioTypeScript120Updated 9 days ago
huggingface/transformersState-of-the-art NLP for everyone — infrastructure for reproducing papersaudiodeep-learningdeepseekgemmaglmhacktoberfestPython168.2k35kUpdated 47 minutes ago
ggml-org/llama.cppRunning LLMs on a laptop — where local experiments beginggmlC++132.4k24.8kUpdated 46 minutes ago
openai/whisperRobust Speech Recognition via Large-Scale Weak SupervisionPython111.3k13.4kUpdated 32 minutes ago

Frontier Radar

Say Hello

Scan to discuss ML and systems