William Shen
Founder CEO at Apparate Labs
Palo Alto, California, United States
Summary
🤩
Rockstar🎓
Top SchoolWilliam Shen is a founder and PhD-trained computer scientist specializing in computer vision, graphics, generative models, and robotics, with 11 years of experience building systems that perceive and act in 3D physical environments. As Founder & CEO of Apparate Labs he is developing real-time generative humans for visual embodiment, building on a Stanford research portfolio guided by Silvio Savarese and Leonidas Guibas and award-winning work (CVPR Best Paper, RSS recognitions). His contributions span research and engineering: from scene conversion and object-loading improvements in the iGibson simulator to data-loading and task configuration fixes in the influential Taskonomy project. He has driven industry research at NVIDIA and Waymo—publishing on deformable object manipulation and realistic simulation—and his NVIDIA work led to a patent filing. Based in Palo Alto, he blends deep academic rigor with product-focused execution, uniquely combining implicit-models research with practical simulation pipelines.
10 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, Doctor of Philosophy - PhD, Computer Science at Stanford University
Github Skills (16)
data-preprocessing10
urdf10
data-loading10
computer-vision10
machine-learning10
dataprep10
preprocessing10
python10
load-data10
image-processing10
preprocess10
tensorflow9
simulations9
simulation9
json8
Programming languages (5)
Github contributions (5)
StanfordVL/iGibson
Aug 2020 - Feb 2021
A Simulation Environment to train Robots in Large Realistic Interactive Scenes
Role in this project:
Back-end Developer
Back-end DeveloperContributions:188 commits in 5 months
Contributions summary:William primarily contributed to scene conversion and object loading functionalities within the iGibson simulation environment. Their commits focused on modifying the scene converter script to handle new data formats, filter objects, and integrate models from different sources like 3D-FRONT. The user also made changes to the interactive objects and scene loading processes, ensuring that objects were correctly scaled, positioned, and loaded. These changes facilitated the integration of external datasets and improvements in the scene visualization pipeline.
roboticssimulationrobot-simulatorrealistic3d-scenes
StanfordVL/taskonomy
Mar 2018 - Mar 2018
Taskonomy: Disentangling Task Transfer Learning [Best Paper, CVPR2018]
Role in this project:ML Engineer
ML EngineerContributions:5 commits, 2 pushes, 4 comments in 3 days
Contributions summary:William primarily focused on bug fixes and improvements to the taskbank/lib/data/load_ops.py file, indicating involvement in the data loading and preprocessing pipeline. Their commits addressed issues related to colorization, jigsaw puzzle tasks, and room layout tasks, suggesting work across different computer vision tasks within the Taskonomy project. The user also modified configuration files related to the colorization and jigsaw tasks, which implies involvement in experiment setup and model configuration.
pytorchdeep-learningtransfer-learningcvpr2018task-transfer
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.