Kaisheng Yao
Software Engineer at Google
Bellevue, Washington, United States
Summary
👤
Senior🎓
Top SchoolKaisheng Yao is a software engineer and team lead at Google with 11 years of experience building foundation models and production ML systems across top-tier research and industry labs. He previously served as Principal Applied Scientist and senior manager at AWS AI, led engineering at Ant Group, and spent a decade-plus at Microsoft Research and labs focusing on speech recognition and language understanding. His background blends deep research (Ph.D. in Communication and Information Systems from Tsinghua) with hands-on systems work—from embedded speech engines at Texas Instruments to open-source contributions to the Microsoft CNTK deep learning toolkit (including CRF sequence objectives and sequence decoding). Based in Bellevue, WA, he has a strong track record of turning advanced algorithms into scalable products and mentoring cross-functional teams. Notably, his career shows a persistent focus on robust speech and sequence modeling across noisy, multilingual settings, a thread connecting his academic and industrial impact.
11 years of coding experience
20 years of employment as a software developer
Ph.D., Communication and Information Systems, Ph.D., Communication and Information Systems at Tsinghua UniversityHong Kong University of Science and Technology (HKUST)
English, Chinese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (12)
neural-network10
machine-learning10
c-language10
cntk10
deep-learning10
cprogramming-language10
maths9
deep-neural-networks9
math9
mathics9
mathematics9
python6
Programming languages (1)
Github contributions (5)
microsoft/CNTK
Sep 2014 - Jul 2015
Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:586 commits in 10 months
Contributions summary:Kaisheng contributed to exposing a log-add math operation and incorporated a CRF sequence-level objective function, indicating work on core algorithms and model training. The user also developed sequence decoding functionality, and implemented a class-based entropy loss function, suggesting they were involved in both model development and algorithm implementation within the deep learning framework. The contributions include implementing and optimizing core mathematical operations and model-specific layers.
pytorchpythondeep-learningc-plus-plusmachine-learning
kaishengyao/keras
May 2015 - May 2015
Contributions:4 commits in 4 days
deep-learningrecurrent-neural-networkstheanoneural-networksmachine-learning
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.