Xingyuan Bu
Senior Algorithm Engineer at Alibaba Group
Hangzhou City, Zhejiang, China
Summary
👤
Senior🎓
Top SchoolXingyuan Bu is a Senior Algorithm Engineer with 11 years of experience combining academic rigor and industry impact in computer vision, multimodality, and large language models. He led teams at Alibaba to scale and fine-tune 13B–65B models with 300K+ instruction examples, outperforming top benchmarks and deploying commercial LLMs for e-commerce use cases. His multimodal and retrieval innovations—including a multi-stage CLIP pipeline and HNSW-based recall—directly lifted ad relevance and generated over a billion RMB in annual revenue. A prolific contributor to PaddlePaddle, he implemented core object-detection ops and data pipelines used in production, reflecting deep end-to-end system skills from operator-level code to model deployment. He brings a rare blend of research publications (CVPR, TPAMI), product-driven results, and hands-on open-source engineering that accelerates both model accuracy and system throughput.
11 years of coding experience
3 years of employment as a software developer
M.S. in Computer Science Computer Vision & Deep Learning, M.S. in Computer Science Computer Vision & Deep Learning at Beijing Institute of Technology
Chinese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (16)
neural-network10
paddlepaddle10
object-detection10
faster-rcnn10
computer-vision10
machine-learning10
deeplearning-ai10
deep-learning10
mscoco10
coco10
python10
modeling10
tensorflow9
cprogramming-language9
c-language9
Programming languages (8)
Github contributions (5)
PaddlePaddle/models
Apr 2018 - Sep 2018
Officially maintained, supported by PaddlePaddle, including CV, NLP, Speech, Rec, TS, big models and so on.
Role in this project:
ML Engineer
ML EngineerContributions:10 commits, 11 PRs, 44 comments in 5 months
Contributions summary:Xingyuan primarily worked on integrating the COCO dataset for object detection using the SSD model within the PaddlePaddle framework. Their contributions include implementing a COCO data reader, modifying training scripts, and evaluating the model's performance with COCO mAP. The user made significant changes to the training pipeline, including learning rate adjustments and model saving. They also added inference capabilities and updated documentation.
nlpdeep-learningpaddlepaddlerecommendationcomputer-vision
PaddlePaddle/Paddle
Jul 2018 - Oct 2018
PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice （『飞桨』核心框架，深度学习&机器学习高性能单机、分布式训练和跨平台部署）
Role in this project:ML Engineer
ML EngineerContributions:7 commits, 5 PRs, 65 comments in 3 months
Contributions summary:Xingyuan implemented and added the FasterRCNN anchor generator operator, including the operator's core logic, unit tests, and Python API. They also added a proposals generation operator and generate proposal labels for the Faster RCNN model. These contributions involved the development of core components within the PaddlePaddle framework, specifically for object detection tasks.
pytorchpythonparalleldeep-learningpaddlepaddle
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.