👋🏻 Hey,

I am Fenlyin.

I build model engines, and make LLMs cheaper to run

• I write Python and PyTorch every day, and I care a lot about what happens underneath a forward pass.
  
• Model engines — attention kernels, fused operators, memory planning, and the scheduler that keeps the GPU busy.
  
• Quantization — GPTQ, AWQ, SmoothQuant, W8A8 / INT4 / FP8, trading a little accuracy for a lot of memory and latency.
  
• Deployment — serving with vLLM and TensorRT-LLM, continuous batching, then Docker and Kubernetes once it has to survive production traffic.
  

My Tech Stack

Python	C++	Java
Git	PyTorch	vLLM

My Bilibili Home

My GitHub stats

THAT'S ME !

⭐️ From Fenlyin0420(mks)