blog | Henry Zhu
maknee.github.io · 410 words · saved by 1 readers
Personal website for some random tidbits I work on
DeepSeek 3FS Performance Journal A series analyzing the filesystem (3FS) DeepSeek uses for LLM training and inference Introduces 3FS, its claims and high‑level design Reviews DeepSeek’s published graphs against theoretical limits Builds and evaluates network and storage microbenchmarks ~ asking how does it perform in practice Highlighted Posts Describes the process of building an end‑to‑end pipeline for scraping/decoding League of Legends replays. Datasets: 700K+ and 700K+ (S12). Intercepts OpenGL pipelines and re‑renders them with NVIDIA ray tracing (and its many various…
saved by
related reading
- Network and Storage Benchmarks for LLM Training on the Cloud | Henry Zhumaknee.github.io
- As Rocks May Think | Eric Jangevjang.com
- How to Land a Frontier Lab Jobvladfeinberg.com
- LLM Visualizationbbycroft.net
- Together AI | The AI Native Cloudtogether.ai
- 2025: The year in LLMssimonwillison.net
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Modelsarxiv.org
- KernelBench: Can LLMs Write GPU Kernels?scalingintelligence.stanford.edu
- Langfuselangfuse.com
- Machine Learning System Resources | std::bodun::blogbodunhu.com
- Latent.Space | Substacklatent.space
- GitHub - wafer-ai/gpu-perf-engineering-resources: A curated resource list for learning AI performance engineering, from GPU fundamentals to production inference.github.com