Wenxin Hou
Applied Scientist 2 at Microsoft
China
Summary
🤩
Rockstar🎓
Top SchoolWenxin Hou is an Applied Scientist with nine years of experience bridging speech and search relevance systems, currently working on relevance and click modeling at Bing Search. Trained at Tokyo Institute of Technology and Nanjing University, Wenxin has hands-on research and engineering experience in speech recognition, transfer learning, and domain adaptation from internships and research roles at Microsoft and Trip.com. An active contributor to prominent open-source projects like ESPnet, Wenxin focuses on making end-to-end speech pipelines robust and maintainable—improving run scripts, dataset handling, and cross-lingual adapter integrations. Comfortable across ML research and production DevOps, Wenxin blends algorithmic work (MMD, adversarial training) with practical engineering fixes and code hygiene to move models from experiments into reliable runs.
9 years of coding experience
Master of Engineering - MEng, Master of Engineering - MEng at Tokyo Institute of TechnologyBachelor of Engineering - BE, Bachelor of Engineering - BE at Nanjing University
Chinese, English, Japanese
Github Skills (15)
pytorch10
transfer-learning10
deeplearning-ai10
bash10
speech-recognition10
machine-learning10
deep-learning10
adaptation10
python10
adversarial-learning9
self-supervised-learning7
machine-translation7
unsupervised-learning7
docker4
dockers4
Programming languages (4)
Github contributions (5)
jindongwang/transferlearning
Apr 2021 - Nov 2021
Transfer learning / domain adaptation / domain generalization / multi-task learning etc. Papers, codes, datasets, applications, tutorials.-迁移学习
Role in this project:
ML Engineer
ML EngineerContributions:23 commits, 7 PRs, 11 comments in 7 months
Contributions summary:Wenxin primarily contributed to a transfer learning project, focusing on adapting and implementing various machine learning models. Their work involved integrating adapter modules for cross-lingual ASR and developing methods for domain adaptation and generalization using techniques like MMD and adversarial training. The commits show active involvement in modifying and extending the existing codebase to incorporate new loss functions and architectural changes. The user also seems to be involved in integrating and testing the developed components within the overall ESPnet framework.
multi-task-learningtransfer-learningtransferlearningdomain-adaptationsurvey
espnet/espnet
May 2020 - Jul 2020
End-to-End Speech Processing Toolkit
Role in this project:Back-end Developer / DevOps Engineer
Back-end Developer / DevOps EngineerContributions:9 commits, 5 PRs, 10 comments in 1 month
Contributions summary:Wenxin primarily contributed to the `espnet/espnet` repository by modifying the `run.sh` scripts, which are likely used for training and decoding speech models. These changes involved adjusting data loading, dataset configurations, and recognition set definitions to include new datasets. The user also fixed a bug related to language ID handling and made improvements to the codebase by applying black formatting. This suggests a focus on the build and execution of the speech processing pipelines and ensuring the code's maintainability.
speech-processingdeep-learningend-to-endchainerpytorch
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.