“士之读书治学，盖将以脱心志于俗谛之桎梏，真理因得以发扬。” —— 陈寅恪

“Science is not the exclusive province of experts and scholars; it is within everyone’s reach, and whoever undertakes it should arrive at the same results.” —— Senku Ishigami

About [Mathew Shen]¶

• 🔭 Currently, I'm a Senior Machine Learning Engineer at NetEase Fuxi, focusing on Data-Centric AI.
• 🔭 Previously, I worked at ByteDance as an Machine Learning Engineer in the AIOps field.
• 🔭 Prior to ByteDance, I contributed to algorithm research in network security (risk management) at Tencent.
• 🔭 My academic background is in statistics, having earned both my undergraduate degree from Hainan University (HNU) and my graduate degree from Sun Yat-sen University (SYSU) in this field.

Projects¶

• 🎓Learning & Eduation

  • 🐍ai glimpse: toyllm, toynlp, toydl, toyml, toyrl
  • 🐍beer: Challenging Problems in Probability and Statistics
  • 🐍oh-numpy, oh-pytorch: Numpy & PyTorch Exercises
  • 🐍toydata: Data Structures in Python

• 🔨Tools

  • 🐍zhplot: 一行代码搞定 Python 图表中文展示
  • 🐍mppt: A Modern Python Package Template
  • 🐍pysesd: Time Series Seasonal Extreme Studentized Deviate(S-ESD) in Python
  • 🦀bleuscore: BLEU Score in Rust
  • 🦀shapley: Shapley Value in Rust
  • 💧qwen: Elixir ❤️ Qwen
  • ✒️ctexbook-template: 一个不太复杂的中文书籍 LaTeX 模板
  • ✒️presentia: Elegant presentation template in LaTeX and Typst

• 🎉Fun

  • ⚛️llm literature: calvino

Blog¶

• 2025-05: LLM KV Cache: A Simple Implementation
• 2025-03: LLM Speculative Sampling
• 2025-02: Deepseek GRPO 中的 KL Divergence
• 2025-02: 大语言模型与深度学习书籍推荐
• 2024-12: LLM in 2024
• 2024-11: 斯坦福小镇 (AI-Town) 系统解读
• 2024-08: Dawid-Skene 算法

Presentation¶

• 2025/02: Python算法库构建实践——复杂必须存在于某处
• 2023/12: MPPT: A Modern Python Package Template
• 2023/08: Data Mining Odyssey
• 2020/09: Technical debt in machine learning systems