Naoyuki Kanda
AI Research Scientist at Meta
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolNaoyuki Kanda is an AI research scientist with 15+ years in spoken language technology and nine years of hands-on deep learning engineering, now based in Seattle and currently at Meta after senior research leadership roles at Microsoft and Hitachi. He combines top-tier academic credentials (PhD, Kyoto University) and first-author publications in IEEE/ICASSP/INTERSPEECH with prize-winning systems in IWSLT, CHiME and VoxSRC competitions, reflecting both theoretical depth and competitive impact. Practically oriented, he has contributed substantial ASR pipeline improvements to the widely used ESPnet toolkit—enhancing tokenization, LM integration, and data prep for speaker-aware scoring. Colleagues know him for bridging rigorous research with production-ready tooling, especially in speech recognition, separation, diarization and TTS. He also brings program-level experience as a reviewer, area chair and associate editor, offering broad community stewardship beyond individual projects.
9 years of coding experience
19 years of employment as a software developer
Doctor of Philosophy (Ph.D.) Informatics, Doctor of Philosophy (Ph.D.) Informatics at Kyoto University
Japanese, English
Github Skills (7)
sh10
bash10
script10
shell10
asr10
scripting10
language-modeling8
Programming languages (6)
Github contributions (5)
espnet/espnet
Oct 2018 - Oct 2022
End-to-End Speech Processing Toolkit
Role in this project:
Back-end Developer
Back-end DeveloperContributions:230 reviews, 1679 commits, 759 PRs in 4 years
Contributions summary:Naoyuki's commits primarily involved modifications to the asr.sh script and related utilities, focusing on enhancements to the data preparation and decoding stages. The changes introduce functionality for incorporating text cleaning and tokenization, including options for BPE and character-level tokenization, and the integration of language model paths. Additionally, they modified the generation and utilization of datasets to provide speaker id and clean speech file path to audio-related scoring processes and created the related test scripts for it. These modifications suggest a focus on enhancing the ASR pipeline.
speech-recognitionspeech-separationchainerspoken-language-understandingspeech-processing
kamo-naoyuki/ctc-segmentation
Aug 2020 - Aug 2022
CTC segmentation python package
Contributions:4 PRs, 30 pushes, 5 branches in 2 years
ctcpython-packagepythonsegmentation
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.