Ye Jia
Research Scientist at Google DeepMind
Mountain View, California, United States
Summary
👤
Senior🎓
Top SchoolYe Jia is a research scientist and entrepreneur with 10 years of experience building state-of-the-art speech and voice systems across industry-leading labs and startups. He has led foundational work at Google Brain on speech synthesis and translation (including Translatotron and zero-shot voice cloning), co-founded Tomato.ai to deliver real-time accent softening and noise cancellation, and contributed to large models and product launches at Meta and OpenAI. Currently at DeepMind, he focuses on advancing speech research with an uncommon blend of production-grade engineering and academic rigor. An active contributor to TensorFlow Lingvo, he has improved tokenization and data-loading pipelines that enable language-dependent processing at scale—evidence of his attention to both model capability and infrastructure. Based in Mountain View, he pairs dual BS degrees from Peking University in mathematics and physics with a track record of shipping research into real-world products.
10 years of coding experience
7 years of employment as a software developer
B.S., Mathematics, B.S., Mathematics at Peking University
Github Skills (13)
machine-translation10
speech-recognition10
nlp10
tensorflow10
python10
data-pipelines9
distributed-computing9
data-pipeline9
speech-to-text9
tokenizer8
tokenize8
language-modeling8
asr8
Programming languages (3)
Github contributions (5)
tensorflow/lingvo
Nov 2018 - Apr 2022
Lingvo
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:42 commits in 3 years 5 months
Contributions summary:Ye primarily contributed to the `lingvo` project, which is a machine learning framework, by enhancing tokenization functionalities to support language-dependent processing and improving various components related to the input generation, including the addition of a padding value argument. The user also exposed outputs from the middle layers of the AsrEncoder and fixed the extra output for conv_lstm in asr/encoder. Further contributions included supporting sharded file patterns and a wildcard character for the number of shards in record yielder, suggesting improvements to data loading and processing pipelines.
asrtranslationctcspeech-recognitiontensorflow
google-research-datasets/cvss
Jan 2022 - Aug 2022
CVSS: A Massively Multilingual Speech-to-Speech Translation Corpus
Contributions:14 commits, 10 pushes, 1 branch in 7 months
nlpspeech-translationtranslationcorpusspeech-recognition
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.