Xingyi Zhou
Scientist at AMI - Advanced Machine Intelligence
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolXingyi Zhou is a research scientist with 11 years of experience applying deep learning to real-world perception problems, currently working at AMI after research roles at Meta and Google. He led development of influential open-source projects like CenterNet/CenterTrack, contributing core model decoding, tracking refinements, and dataset support that helped shape modern object detection and multi-object tracking workflows. With a PhD from UT Austin and intern experience across top labs (Facebook AI, Microsoft Research Asia, Intel), he blends rigorous research with production-minded engineering—optimizing for FP16, memory, and Detectron2 compatibility. His work shows a consistent focus on making state-of-the-art models robust and usable, from visualization and demo refactors to experiment and evaluation tooling. Based in New York, he is known for pragmatic code refactors and cross-project maintenance that keep research systems shipping.
12 years of coding experience
4 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at The University of Texas at AustinBachelor of Science (BS), Computer Science, Bachelor of Science (BS), Computer Science at Fudan University
Chinese
Github Skills (17)
pytorch10
python10
evaluation10
machine-learning10
detectron10
eval10
trainings10
object-detection10
computer-vision10
modeling10
model-optimization9
experiment9
experiment-manager9
tensorflow8
git8
Programming languages (3)
Github contributions (5)
xingyizhou/CenterNet
Apr 2019 - Jun 2020
Object detection, 3D detection, and pose estimation using center point detection:
Role in this project:
ML Engineer
ML EngineerContributions:31 commits, 21 PRs, 26 pushes in 1 year 2 months
Contributions summary:Xingyi primarily contributed to the model definition and evaluation scripts of CenterNet. The commits show modifications to the `decode.py` file, indicating involvement in the core object detection decoding process, including the implementation of different aggregation functions. Additionally, the user added, modified, and maintained experiment scripts and evaluation functions within the project. These modifications suggest a focus on training and evaluating object detection models.
3d-detectionobject-detectionpose-estimation
xingyizhou/CenterNet2
Mar 2021 - Apr 2022
Two-stage CenterNet
Role in this project:ML Engineer
ML EngineerContributions:20 commits, 16 pushes, 60 comments in 1 year 1 month
Contributions summary:Xingyi primarily focused on modifying and enhancing the CenterNet2 model, a two-stage object detection system. Their contributions included refactoring code, removing unused elements, and adding references within the codebase. A significant portion of their work involved adapting the code for compatibility with updated Detectron2 versions, indicating a focus on maintaining and improving the project's core functionality. Further work involved integrating support for new datasets such as nuImages, and debugging and optimizing the code for FP16 compatibility and memory usage.
pytorchcocoobject-detectioncenternetstage
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.