- π Based in: Shanghai, China
- π’ Role: AI Infra Team at π Xiaohongshu
- π€ Focus: Post-training of large language models (SFT, RL, alignment, optimization)
- π» Stack: Python, PyTorch, Distributed Training, Optimization; Golang, C++, C#, TeX, Elispβ¦
- π¬ MLLM post-training techniques, especially large-scale reinforcement learning with multimodal data
- β‘ Training/Inference acceleration & model efficiency
- 𧩠Improving stability and scalability of AI infrastructure
;; This is my work...
(=> (++ (βοΈ π β‘) (π§ π π))
(=> (++ π π¦)
(π π)))- β¨ Deep Emacs enthusiast β I write, organize, and live in Emacs + Org Mode
- β Pour-over coffee lover β exploring beans, refining techniques, enjoying the process with
COMANDANTE::C40. - π΅ Post-rock music
β Smooth is π fast
See how to do the nyan on NYAN.CAT!