Yezhen Cong
Software Engineer IV at Databricks
Mountain View, California, United States
Summary
🤩
Rockstar🎓
Top SchoolYezhen Cong is a Software Engineer IV based in Mountain View with six years of experience bridging cutting-edge computer vision research and scalable backend engineering, currently driving Unity Catalog performance and scale at Databricks. A Tsinghua Outstanding Graduate and Stanford MSCS alum, he has a strong publication record (joint-first authors at CVPR and NeurIPS) and a history of shipping research code into production-quality toolkits like OpenMMLab’s mmdetection3d. His work spans implementing novel 3D detection losses and GPU-accelerated ops to improving AI metadata and data lineage systems, showing comfort across ML infrastructure and model-level contributions. Recognized as a top course assistant at Stanford, he combines rigorous academic training with pragmatic engineering—often refactoring large codebases single-handedly to make advanced research reproducible and performant.
6 years of coding experience
4 years of employment as a software developer
No. 2 High School of East China Normal UniversityBachelor of Engineering - BE, Computer Software Engineering, 3.98/4.0, Bachelor of Engineering - BE, Computer Software Engineering, 3.98/4.0 at Tsinghua UniversityMaster of Science - MS, Computer Science, 4.154/4.3 (4.3 is A+), Master of Science - MS, Computer Science, 4.154/4.3 (4.3 is A+) at Stanford University
Github Skills (6)
object-detection10
computer-vision10
pytorch10
3d-object-detection10
point-cloud9
cuda9
Programming languages (1)
Github contributions (5)
open-mmlab/mmdetection3d
Nov 2020 - Nov 2021
OpenMMLab's next-generation platform for general 3D object detection.
Role in this project:
ML Engineer
ML EngineerContributions:169 reviews, 54 commits, 72 PRs in 1 year
Contributions summary:Yezhen implemented and integrated axis-aligned IoU loss into the VoteNet model for 3D object detection. This involved adding new functionality to calculate IoU, modifying the VoteHead and bbox coder, and incorporating the loss into the training process. The changes involved code modifications in core modules and model heads. Furthermore, the user fixed bugs caused by mmcv upgrades and added support for a GPU-accelerated KNN operation.
pytorchnext-generationkitti-datasetdeep-learningobject-detection
yezhen17/3DIoUMatch
Nov 2020 - Jun 2022
[CVPR 2021] PyTorch implementation of 3DIoUMatch: Leveraging IoU Prediction for Semi-Supervised 3D Object Detection.
Contributions:38 commits, 1 PR, 32 pushes in 1 year 6 months
pytorchsupervisedpredictionleveragingiou
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.