季威
Data Mining Engineer Team Manager
Haidian District, Beijing, China
Summary
🤩
Rockstar🎓
Top School季威 is a data mining engineering leader with 11 years of experience building large-scale data pipelines and relevance systems, currently managing a team focused on training and scaling AI foundation models at StepFun in Beijing. He spent eight years at Microsoft leading Bing Whole Page Relevance efforts, where he combined data mining, pipeline engineering, and search experience optimization as a tech lead. Early work at Baidu on video vertical search and active contributions to computer-vision research repos (e.g., implementations of ICCV2019 LearningToPaint and ECCV2022 RIFE) show a strong ML/vision applied research bent beyond pure infra. Comfortable shipping production systems and tweaking model architectures, he brings a pragmatic balance of research-minded experimentation and production-grade engineering. Collected in Beijing University of Posts and Telecommunications, his background blends deep search expertise with hands-on ML model and GPU-centered engineering.
11 years of coding experience
9 years of employment as a software developer
Bachelor’s Degree, Computer Software Engineering, Bachelor’s Degree, Computer Software Engineering at Beijing University of Posts and Telecommunications
Chinese
Github Skills (13)
mask-rcnn10
faster-rcnn10
pytorch10
computer-vision10
deeplearning-ai10
deep-learning10
trainings10
python10
reinforcement-learning10
modeling10
resnet9
image-generation8
tensorflow4
Programming languages (5)
Github contributions (5)
hzwer/ECCV2022-RIFE
Nov 2020 - Jan 2023
ECCV2022 - Real-Time Intermediate Flow Estimation for Video Frame Interpolation
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 534 commits, 34 PRs in 2 years 2 months
Contributions summary:季威 contributed to the development and implementation of a video frame interpolation model, adding and modifying code for the core model. They introduced a new model based on Real-Time Intermediate Flow Estimation (RIFE) for video frame interpolation, indicated by changes to model architecture files, including Flownet and the addition of contextual and fusion networks. The user focused on integrating and refining deep learning models for computer vision tasks.
framedeep-learningoptical-flowcomputer-graphicscomputer-vision
hzwer/ICCV2019-LearningToPaint
Mar 2019 - Sep 2022
ICCV2019 - Learning to Paint With Model-based Deep Reinforcement Learning
Role in this project:ML Engineer
ML EngineerContributions:93 commits, 11 PRs, 163 pushes in 3 years 7 months
Contributions summary:季威 primarily contributed to the codebase by modifying and updating training scripts, model definitions, and test code. Their changes include adjusting file names, merging code, cleaning up codes, updating model parameters, and fixing issues related to GPU usage and model saving. The user also added code to generate and save images, suggesting a focus on the visual output of the painting models. This aligns with the project's goal of learning to paint using deep reinforcement learning.
pytorchmodel-baseddeep-learningpaintingreinforcement-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.