Siyuan Qiao
Research Scientist at Meta
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolSiyuan Qiao is a research scientist with nine years of experience building cutting-edge AI systems across Google, DeepMind, ByteDance, and Meta, and holds a PhD in Computer Science from Johns Hopkins. He was a founding member of Gemini at DeepMind and contributed to Gemini 1–3, later working on large-scale pretraining at ByteDance before joining Meta to focus on theory-driven determinism in ML. Siyuan blends deep research with practical engineering—evidenced by contributions to Google Research’s DeepLab2 where he implemented a DVPS data-building pipeline that integrates depth into video panoptic segmentation workflows. Based in Mountain View, he moves fluidly between model research, data infrastructure, and production-ready tooling, with a track record of shipping components used in state-of-the-art vision and language systems. Colleagues describe him as someone who pairs theoretical rigor with hands-on code, often surfacing non-obvious dataset and preprocessing optimizations that accelerate model development.
9 years of coding experience
4 years of employment as a software developer
Johns Hopkins University
Github Skills (8)
data-preprocessing10
computer-vision10
deep-learning10
tensorflow10
python10
segmentation9
data-pipeline9
image-segmentation9
Programming languages (2)
Github contributions (5)
google-research/deeplab2
Jun 2021 - Jul 2021
DeepLab2 is a TensorFlow library for deep labeling, aiming to provide a unified and state-of-the-art TensorFlow codebase for dense pixel labeling tasks.
Role in this project:
ML Engineer
ML EngineerContributions:5 reviews, 29 commits, 3 PRs in 1 month
Contributions summary:Siyuan's primary contribution involves adding a data building tool specifically for the DVPS (Depth-aware Video Panoptic Segmentation) task within the DeepLab2 project. This tool likely prepares the DVPS dataset for training, enabling the integration of depth information alongside images and segmentation labels. The changes include modifications to the `build_dvps_data.py` and `data_utils.py` files, indicating a focus on data loading and preprocessing for this specific computer vision task. The commits demonstrate direct interaction with the project's data processing pipeline.
artdeep-learningtensorflow-librarystate-of-the-artaiming
joe-siyuan-qiao/ViP-DeepLab
Nov 2020 - Jul 2021
Contributions:11 commits, 2 pushes, 9 comments in 7 months
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.