一個 AI compiler Engineer 需要的技能
以反推的方式在 2024 年夏季出現在各個職缺應徵的 Job Descript 裡列出來所需要的條件
扮演的角色及任務
- 建立客製化 MLIR-based compiler
- 將 AI model 轉成不同 device 架構上的 instruction
- 實作 optimization 並部署在不同的 framework 或 device
所需的技能及條件
- machine learning 相關 compiler:
- Triton, XLA, Glow, TVM, nGraph, TensorRT, IREE, etc.
- ML Graph optimizations
- C++ 深度理解
- Cmake, GGC
- MLIR, LLVM
- 8 bit and lower model quantization
- GPU profiling
- nsys
- nsight
延伸名詞跟脈絡
- 近代有哪些 ML Graph Optimization 做法跟問題
- MLIR Base 跟 LLVM 的架構跟練習
- 有哪些 framework
- 有哪些 target device 架構
https://mlc.ai/
https://github.com/zwang4/awesome-machine-learning-in-compilers
https://www.reddit.com/r/Compilers/comments/14z6gwd/any_recommandation_for_studying_ai_compilers/
https://huyenchip.com/2021/09/07/a-friendly-introduction-to-machine-learning-compilers-and-optimizers.html
https://arxiv.org/abs/2101.04808
Reference
https://www.modular.com/career-post?4287507005&gh_jid=4287507005
https://www.linkedin.com/jobs/view/ai-compiler-engineer-at-mediatek-3874407229/?originalSubdomain=tw
https://www.cakeresume.com/companies/skymizer-taiwan-inc/jobs/senior-software-engineer-ai-compiler?locale=en