Wang Peng
Beijing, China
Summary
🤩
RockstarWang Peng is a research scientist with 8 years of experience focused on deep learning and vision-language models, currently contributing to the Qwen team at Alibaba in Beijing. He has hands-on expertise in model development, evaluation, and deployment, evidenced by substantive contributions to the well-known OFA (One-For-All) ICML 2022 repository where he enhanced training/evaluation workflows and added zero-shot refcoco capabilities. Comfortable working at the intersection of research and engineering, he bridges algorithmic innovation with production-ready tooling and code refactoring. Known for pragmatic problem solving, he often improves model evaluation pipelines to enable more reliable downstream benchmarks. Based in China, he combines strong open-source engagement with large-scale industrial research experience.
8 years of coding experience
Github Skills (9)
computer-vision10
pytorch10
machine-learning10
pre-trained-model9
natural-language-processing9
image-annotation9
image-text9
nlp9
multimodal9
Programming languages (3)
Github contributions (5)
OFA-Sys/OFA
Feb 2022 - Jan 2023
Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework
Role in this project:
ML Engineer
ML EngineerContributions:4 reviews, 166 commits, 61 PRs in 11 months
Contributions summary:Wang primarily contributed to the OFA (One-For-All) model, focusing on vision-language tasks within the repository. Their work involved updating and modifying the core `ofa_task.py` file, which likely manages the overall training and evaluation process. The user also worked on implementing and refining various scripts for evaluating models in caption and refcoco tasks, in addition to adding refcoco zero-shot evaluation and relevant code refactoring. These contributions suggest an emphasis on model deployment and evaluation.
sequence-to-sequencemultimodalpretrainingimage-captioningtext-to-image-synthesis
OFA-Sys/ONE-PEACE
May 2023 - Jun 2025
A general representation model across vision, audio, language modalities. Paper: ONE-PEACE: Exploring One General Representation Model Toward Unlimited Modalities
Contributions:3 PRs, 93 pushes, 30 branches in 2 years 1 month
audiofoundation-modelsmultimodalrepresentation-learningvision-language
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.