Junbo Zhang
Senior Researcher at Xiaomi Technology
Beijing, China
Summary
🤩
Rockstar🎓
Top SchoolJunbo Zhang is a senior researcher and engineer with 11 years of experience specializing in speech recognition and synthesis, currently based in Beijing and working at Xiaomi. He holds a Ph.D. in Signal and Information Processing and has applied academic rigor to industry projects at Samsung and LeEco before joining Xiaomi. Junbo is an active open-source contributor, notably enhancing Kaldi with Goodness of Pronunciation (GOP) functionality and improving tiny-dnn’s CBLAS-backed fully connected ops for performance gains. He also contributed cross-platform tooling and example improvements to lhotse, demonstrating a practical focus on reproducible speech datasets and developer experience. Known for backend and DevOps fluency, he bridges low-level C++ optimization with higher-level ML data workflows. An interesting detail: his contributions span both research-grade algorithms for mispronunciation detection and pragmatic engineering work that makes speech toolchains easier to build and run across platforms.
11 years of coding experience
9 years of employment as a software developer
Ph.D., Signal and Information Processing, Ph.D., Signal and Information Processing at University of Chinese Academy of SciencesBachelor, Electronic and Communication, Bachelor, Electronic and Communication at Beijing University of Chemical Technology
English, Chinese
Github Skills (30)
cblas10
c-language10
python10
scripting10
audio10
kaldi10
script10
deep-learning10
speech10
cplus10
neural-network10
sh10
speech-recognition10
cpp10
shell10
Programming languages (6)
Github contributions (5)
lhotse-speech/lhotse
Jun 2020 - Aug 2021
Tools for handling speech data in machine learning projects.
Role in this project:
Backend Developer & DevOps Engineer
Backend Developer & DevOps EngineerContributions:19 reviews, 72 commits, 16 PRs in 1 year 2 months
Contributions summary:Junbo's primary contributions focused on improving the examples within the project, particularly the `librimix` example. Their commits implemented changes to the build configuration to ensure it functions correctly on macOS and also refactored the file paths to reside within a `data` directory instead of `env`. Moreover, they were responsible for fixing various aspects of the shell scripts and ensuring they are compatible with different platforms.
pytorchkaldipythondatamachine-learning-projects
kaldi-asr/kaldi
Feb 2021 - Feb 2021
kaldi-asr/kaldi is the official location of the Kaldi project.
Role in this project:Back-end Developer
Back-end DeveloperContributions:10 reviews, 2 commits, 8 PRs in 3 days
Contributions summary:Junbo contributed to the Kaldi project by implementing and refining the Goodness of Pronunciation (GOP) feature. This involved writing C++ code to compute GOP from speech probabilities, which is a critical step in mispronunciation detection. The user's work also includes modifications to scripts and example files, demonstrating a focus on integrating the GOP functionality within the existing Kaldi framework and providing practical usage examples.
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.