Zhe Feng

I'm currently a graduate student at National Laboratory of Pattern Recognition (NLPR), Institute of Automation, Chinese Academy of Sciences (CASIA). I am interested in Deep Learning, with a special focus on Computer Vision and Autonomouc Driving. My latest works focus on Efficient Transformer Theory and Vision-based 3D representation learning.

Paper is sooo 😯😦😧😮😲🥱 cheap, I really need representative work 😭😭😭...

Representative Work

	before the code-agent era
12.1K	LeetCUDA — CUDA kernel learning and benchmarking collection. Fixed a numerical-overflow bug in the softmax exponential computation.
7.8K	tensorrtx — popular networks implemented with the TensorRT network definition API. Added GhostNetV1/V2 support.
4.4K	Efficient-AI-Backbones — efficient backbone networks from Huawei Noah’s Ark Lab.
369	tensorrt_starter — tutorial codebase for TensorRT deployment.
	in the code-agent era
6K	flash-linear-attention — efficient implementations of emerging model architectures. Added Mamba-3.
301	nsight-python — Python kernel-profiling interface built on NVIDIA Nsight Tools.
291	Emerging-Optimizers — NVIDIA NeMo research project on emerging optimizers (Shampoo, SOAP, Muon) for LLM training.
84	flash-bidirectional-linear-attention — Triton implementation of bidirectional (non-causal) linear attention. Added RALA.

2025.12 - present, Didi Chuxing Co.Ltd, Beijing, China. Research Intern

2025.02 - 2025.11, Hirain Academia Sinica, Beijing, China. Engineering Intern

2024.09 - 2025.01, School of Vehicle and Mobility, Tsinghua University. Research Intern