Yingruo Fan
Staff Research Scientist at Sony
Mountain View, California, United States
Summary
👤
Senior🎓
Top SchoolYingruo Fan is a Staff Research Scientist based in Mountain View with eight years of experience at the intersection of computer graphics, vision, and deep learning, focused on character animation and photorealistic human synthesis. After earning a PhD from The University of Hong Kong, she progressed through research roles at Samsung Research America and Sony, advancing speech-driven and behavior-driven human animation systems. Her open-source work includes contributions to FaceFormer (CVPR 2022), where she integrated Wav2Vec2 audio encoders and transformer decoders to improve 3D facial motion prediction across datasets. Known for adapting state-of-the-art audio and transformer architectures to practical animation pipelines, she bridges academic research and production-ready ML engineering. Colleagues note her knack for tuning model architectures and data handling to squeeze performance gains in challenging multimodal tasks.
8 years of coding experience
3 years of employment as a software developer
Bachelor's degree, Bachelor's degree at Beijing University of Posts and TelecommunicationsThe University of Hong Kong (HKU)
Chinese, English
Github Skills (10)
transformers10
computer-vision10
pytorch10
deep-learning10
wafv210
speech9
pythorch9
py9
3d-models8
computer-graphics8
Programming languages (1)
Github contributions (5)
EvelynFan/FaceFormer
Mar 2022 - Mar 2022
[CVPR 2022] FaceFormer: Speech-Driven 3D Facial Animation with Transformers
Role in this project:
ML Engineer
ML EngineerContributions:17 commits, 3 pushes, 29 comments in 15 days
Contributions summary:Yingruo's commits primarily focus on modifications and extensions to a deep learning model for speech-driven 3D facial animation. Their work involves incorporating a pre-trained Wav2Vec2 model for audio encoding and integrating transformer decoder layers for motion prediction. Code changes include adjusting network architectures, implementing alignment and periodic positional biases, and modifying data handling for different datasets (BIWI and VOCASET). These modifications suggest an effort to improve model performance and adapt it to varied input data and requirements.
3d-face3d-modelscomputer-graphicscomputer-visiondeep-learning
EvelynFan/AWESOME-MER
Jul 2020 - Oct 2020
Contributions:12 commits, 10 pushes, 1 branch in 2 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.