马宁胜
Yangpu District, Shanghai, China
Summary
🤩
Rockstar🎓
Top School马宁胜 is a computer vision and deep learning engineer with nine years of experience building, training, quantifying, and deploying models, with deep expertise in face recognition pipelines and architectures like ResNet, MobileNet, Faster-RCNN, YOLO, SSD and UNet. He combines strong algorithmic foundations in ML (LR, SVM, RF, clustering methods) and data structures with practical proficiency in PyTorch, TensorFlow, Caffe and deployment toolchains including ONNX and TensorRT. His open-source contributions to OpenMMLab projects (mmdetection, mmdeploy, mmcv) have focused on making object detection models exportable and numerically consistent across PyTorch, ONNX and TensorRT, enabling dynamic shapes and custom NMS/interpolation handling. He also worked on integrating large language models into the OpenCompass evaluation platform, showing versatility beyond vision into LLM tooling and evaluation. Comfortable coding in Python and C++ on Linux and fluent with engineering tools like Git, CMake and GDB, he pairs fast learning and strong self-drive with hands-on deployment experience. Based in Shanghai and holding advanced studies from Fudan and a bachelor’s in mechatronics from Zhejiang University, he brings both theoretical depth and practical system-level know-how.
8 years of coding experience
2 years of employment as a software developer
Master's Degree, Computer Software Engineering, Master's Degree, Computer Software Engineering at Fudan University学士, 机械电子工程, 学士, 机械电子工程 at 浙江大学
English
Github Skills (24)
continuous-deployment10
pytorch10
benchmark10
python10
evaluation10
tensorrt10
benchmarking10
onnx10
ml-deployment10
llm10
deep-learning10
model-optimization10
object-detection10
computer-vision10
model-conversion10
Programming languages (7)
Github contributions (5)
open-mmlab/mmdeploy
Jun 2021 - Jan 2023
OpenMMLab Model Deployment Framework
Role in this project:
ML Engineer
ML EngineerContributions:8 releases, 1150 reviews, 138 commits in 1 year 7 months
Contributions summary:马宁胜's commits focused on supporting the RetinaNet model by adding PyTorch to ONNX conversion. They modified the `onnx_helper.py` file to include dynamic box clipping and a dummy non-maximum suppression (NMS) operator for ONNX export. Further, the user worked on rewriting the `delta2bbox` function and modifying the `SingleStageDetector` and `AnchorHead` classes to improve the model's exportability to ONNX. These changes were targeted at the `mmdeploy` deployment framework.
pytorchncnnpplnndeep-learningdeployment
open-mmlab/mmdetection
Oct 2020 - Nov 2022
OpenMMLab Detection Toolbox and Benchmark
Role in this project:ML Engineer
ML EngineerContributions:155 reviews, 16 commits, 20 PRs in 2 years 1 month
Contributions summary:马宁胜 primarily contributed to improving the exportability of the `mmdetection` models to ONNX and TensorRT for deployment. Their work included modifications to support dynamic input shapes, address reshape errors, and incorporate features such as ONNX simplification. They also focused on enabling specific models, such as YOLOv3 and those in the MaskFormer/DETR series, to work with these deployment tools, while ensuring numerical consistency between PyTorch and ONNX/TensorRT outputs.
retinanetbenchmarkfast-rcnnopenmmlabsemisupervised-learning
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.