Takaaki Saeki
Senior Research Scientist at Google DeepMind
New York, New York, United States
Summary
👤
Senior🎓
Top SchoolTakaaki Saeki is a senior research scientist specializing in spoken language processing, deep learning, and signal processing, currently based in New York and completing a PhD at the University of Tokyo. With eight years of industry and academic experience—including internships and researcher roles at Google, Google DeepMind, Carnegie Mellon, and LINE—he bridges cutting-edge research and practical system integration. He has contributed to major open-source toolkits such as ESPnet, where he integrated and tested TTS models and enabled speaker embedding via xvectors, demonstrating both implementation and data-prep expertise. His background spans speech enhancement, music generation, and text-to-speech, reflecting a rare combination of audio-focused research and production-minded engineering. Notably, he moved from student researcher and JSPS fellow roles into sustained research positions at Google and DeepMind, signaling rapid progression from academia to high-impact industrial research.
8 years of coding experience
4 years of employment as a software developer
University of Tokyo
Github Skills (11)
data-preprocessing10
speech-to-text10
pytorch10
machine-learning10
text-to-speech10
dataprep10
bash9
speech-synthesis9
deep-learning9
python8
kaldi7
Programming languages (3)
Github contributions (5)
espnet/espnet
Aug 2021 - Oct 2022
End-to-End Speech Processing Toolkit
Role in this project:
ML Engineer
ML EngineerContributions:7 reviews, 19 commits, 5 PRs in 1 year 2 months
Contributions summary:Takaaki primarily contributed to the integration and testing of the jtubespeech/tts1 model within the espnet framework. Their work involved implementing scripts for downloading and preparing the JTUBESPEECH dataset, modifying existing scripts for data preparation and splitting, and fixing various bugs. Furthermore, the user updated the run script and enabled the use of xvectors for speaker embedding within the TTS model.
speech-recognitionspeech-separationchainerspoken-language-understandingspeech-processing
sarulab-speech/UTMOS22
Apr 2022 - Nov 2022
UT-Sarulab MOS prediction system using SSL models
Contributions:17 commits, 2 PRs, 9 pushes in 7 months
sslmosprediction
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.