Hui Zhang
Algorithm Expert at 滴滴出行
Haidian District, Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolHui Zhang is an algorithm expert with 10 years' experience building speech generation and recognition systems, currently leading algorithm work at Didi in Beijing. He blends deep research knowledge in TTS, ASR and LLM/Agent technologies with hands-on software and backend engineering, having contributed to flagship open-source projects like Kaldi, PaddlePaddle and ESPnet. His work spans core algorithm development (CTC/RNN-T, beam search, acoustic modeling) to production-grade integrations and GPU-enabled gradient implementations, reflecting both research rigor and pragmatic engineering. At Baidu he was a key contributor to an acclaimed emotional TTS project and led machine-learning defenses deployed at scale, showing a track record of delivering high-impact systems. Known for improving code quality and usability in major speech toolkits, he also brings experience in productization and technical leadership across industry and research settings.
10 years of coding experience
10 years of employment as a software developer
Master's degree, speech recognition and signal processing, Master's degree, speech recognition and signal processing at University of Chinese Academy of SciencesBachelor, Physics, Bachelor, Physics at Tsinghua UniversityMachine learning and deep learning, Machine learning and deep learning at coursera
Github Skills (35)
speech-to-text10
beam-search10
machine-translation10
c-language10
integrate10
speech-processing10
python10
scripting10
testing10
machine-learning10
integrations10
kaldi10
deeplearning-ai10
deep-learning10
tensorflow10
Programming languages (13)
Github contributions (5)
PaddlePaddle/PaddleSpeech
Feb 2021 - Dec 2022
Easy-to-use Speech Toolkit including Self-Supervised Learning model, SOTA/Streaming ASR with punctuation, Streaming TTS with text frontend, Speaker Verification System, End-to-End Speech Translation and Keyword Spotting. Won NAACL2022 Best Demo Award.
Role in this project:
ML Engineer
ML EngineerContributions:5 releases, 1357 reviews, 1351 commits in 1 year 11 months
Contributions summary:Hui's commits primarily focus on modifying and utilizing existing code within the paddlespeech/paddlespeech repository. The changes show an active involvement in integrating and refining existing functionality, specifically concerning speech recognition models. A notable effort is directed towards adapting and integrating various types of models for diverse applications, including model conversion and performance enhancements. These contributions showcase a focus on enhancing the usability and functionality of existing systems within the speech processing toolkit.
voice-recognitionasrtranslationspeech-recognitionstreaming-tts
kaldi-asr/kaldi
Feb 2017 - Mar 2020
kaldi-asr/kaldi is the official location of the Kaldi project.
Role in this project:Backend Developer
Backend DeveloperContributions:9 PRs, 6 comments, 3 issues in 3 years 1 month
Contributions summary:Hui primarily focused on bug fixes and minor improvements to the Kaldi project. Their contributions included addressing compatibility issues related to character encoding in data preparation scripts, correcting error messages, and fixing typos in comments within various source files. They also updated the run script for the Aishell example to show all WERs. The user's work demonstrates a focus on maintaining code quality and improving the usability of the Kaldi toolkit.
cudakaldiasrspeech-to-textkaldi-asr
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.