Jianlin Peng
Senior Applied Scientist at Microsoft
Suzhou City, Jiangsu, China
Summary
🤩
Rockstar🎓
Top SchoolJianlin Peng is a Senior Applied Scientist based in Suzhou with nine years of experience building and deploying computer vision and multimodal ML systems at scale. Currently at Microsoft, he progressed from Applied & Data Scientist to Senior Applied Scientist, bringing production experience from leading Chinese AI companies like SenseTime and Cloudwalk. His open-source contributions include interface and grounding work for Microsoft’s influential UniLM/Kosmos-2 multimodal projects and practical MAE (Masked Autoencoder) refinements that improve pretraining and fine-tuning in PyTorch. Jianlin combines research-grade algorithmic skills with hands-on engineering—implementing normalization tweaks, visualization tools, and Gradio demo apps that bridge models to users. He holds a Master’s in Information Technology from the University of Melbourne and a Mathematics degree from Shanghai Jiao Tong University, a background that helps him connect rigorous theory with robust engineering. Colleagues would describe him as detail-oriented and product-minded, with a knack for turning pretraining insights into usable model features.
9 years of coding experience
3 years of employment as a software developer
The University of MelbourneBachelor's degree, Mathematics, Bachelor's degree, Mathematics at Shanghai Jiao Tong University
Chinese, English
Stackoverflow
1reputation
0reached
1answer
0questions
Github Skills (13)
computer-vision10
pytorch10
parameter-tuning10
machine-learning10
gradio10
nlp10
multimodal10
python10
pre-trained-model10
fine-tuning10
image-processing10
deep-learning9
tensorflow6
Programming languages (2)
Github contributions (5)
pengzhiliang/MAE-pytorch
Nov 2021 - Dec 2021
Unofficial PyTorch implementation of Masked Autoencoders Are Scalable Vision Learners
Role in this project:
ML Engineer
ML EngineerContributions:18 commits, 3 PRs, 16 pushes in 18 days
Contributions summary:Jianlin's contributions primarily involve implementing and refining the Masked Autoencoder (MAE) model within the PyTorch framework. They added the capability to normalize the target patch pixels, a critical modification to the pretraining process. The commits include bug fixes and the implementation of a fine-tuning process. Furthermore, the user integrated visualization capabilities and included weight for the model to assist with the training.
pytorchscalablevisiondeep-learningautoencoders
microsoft/unilm
Sep 2022 - Sep 2022
Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities
Role in this project:ML Engineer
ML EngineerContributions:1 commit, 13 PRs, 103 comments in 1 day
Contributions summary:Jianlin primarily contributes to the `kosmos-2/demo/gradio_app.py` file, indicating development of a user interface for the Kosmos-2 model, a multimodal large language model. Their commits include updates to the Gradio interface, the addition of sampling parameters, and examples. The user also works on integrating grounding features and visualizing model outputs. Further contributions involve preparing and processing data related to image analysis and model training.
layoutxlmtraininglanguage-understandingvision-and-languagewavlm
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.