Xuankai Chang
Machine Learning Researcher at Apple
Pittsburgh, Pennsylvania, United States
Summary
🤩
Rockstar🎓
Top SchoolXuankai Chang is a machine learning researcher with 11 years of experience focused on speech processing and self-supervised representation learning, currently working at Apple after research internships at Google, Microsoft, and Mitsubishi Electric Research Labs. He is pursuing a PhD in Computer Science at Carnegie Mellon University and has a strong academic foundation from Johns Hopkins and Shanghai Jiao Tong University. Xuankai contributes actively to high-profile open-source toolkits like s3prl and ESPnet, where his work on ASR downstream tasks, wav2vec2 interfaces, and multi-speaker models has improved dataset support, augmentation, and evaluation metrics. Comfortable bridging research and engineering, he implements core back-end components and practical fixes that make complex speech models more usable in real systems. Despite joking about limited signal-processing intuition, his contributions show a practical mastery of ML pipelines and reproducible tooling for speech research.
11 years of coding experience
Johns Hopkins UniversityDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Carnegie Mellon UniversityMaster's degree Computer Science, Master's degree Computer Science at Shanghai Jiao Tong University
Github Skills (10)
pytorch10
machine-learning10
speech-recognition10
deep-learning10
automatic-speech-recognition10
python10
asr10
data-processing10
fairseq9
nlp8
Programming languages (5)
Github contributions (5)
espnet/espnet
Jan 2019 - Jan 2023
End-to-End Speech Processing Toolkit
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:333 reviews, 118 commits, 76 PRs in 4 years
Contributions summary:Xuankai's primary contribution appears to be focused on implementing features and making modifications in the `espnet/asr/pytorch_backend/e2e_asr_mix.py` file, indicating their involvement in the core functionalities of the speech processing toolkit. Their work includes adding components related to multi-speaker ASR like the `PIT` class and `EncoderMix`, implementing loss calculations, and computing metrics like WER and CER. The changes also suggest the user is working on supporting MIMIO-speech models.
speech-recognitionspeech-separationchainerspoken-language-understandingspeech-processing
s3prl/s3prl
Feb 2021 - May 2021
Self-Supervised Speech Pre-training and Representation Learning Toolkit
Role in this project:ML Engineer
ML EngineerContributions:5 commits, 3 PRs, 2 branches in 3 months
Contributions summary:Xuankai primarily contributed to the `s3prl/s3prl` repository, which focuses on self-supervised speech pre-training and representation learning. Their commits included modifications to ASR (automatic speech recognition) downstream tasks, including the addition of datasets and spec augmentation support. They updated the wav2vec2 interface due to changes in fairseq and added file locking for downloads. The user also made minor fixes to the ASR dataset, demonstrating their involvement in model development and data processing related to speech pre-training and representation learning.
vq-apcspeech-representationdecoarrepresentation-learningspeech-recognition
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.