Chung-ming Chien
Speech AI Scientist at Toyota Technological Institute at Chicago
Chicago, Illinois, United States
Summary
🤩
Rockstar🎓
Top SchoolChung-ming Chien is a Speech AI Scientist and Ph.D. student at TTIC with seven years of experience advancing speech and NLP systems across academia and industry. His work spans self-supervised speech representations, speech generation, and speech-language models, with publications and benchmarks that reveal intrinsic word-level structure and enable zero-shot spoken language understanding. He has applied these research strengths during internships and roles at Amazon Alexa TTS, Meta (Voicebox), NVIDIA, and Kyutai, where he helped augment LLMs with speech generation and improved controllability and TTS quality. An active ML engineer on GitHub, he enhanced a FastSpeech2 implementation—integrating HiFi-GAN and multi-dataset preprocessing—to produce higher-fidelity samples. Equally comfortable with rigorous theory and production-focused engineering, he brings cross-cultural training from NTU and TTIC and a knack for translating speech research into deployable systems.
7 years of coding experience
5 years of employment as a software developer
Doctor of Philosophy - PhD, Computer Science, 4.0/4.0, Doctor of Philosophy - PhD, Computer Science, 4.0/4.0 at Toyota Technological Institute at ChicagoMaster of Science - MS, Computer Science and Information Engineering, GPA: 4.02/4.3, Master of Science - MS, Computer Science and Information Engineering, GPA: 4.02/4.3 at National Taiwan University
Chinese, English, German, Mandarin
Github Skills (7)
pytorch10
machine-learning10
speech-synthesis10
tensorboard9
data-processing8
python8
nlp7
Programming languages (2)
Github contributions (5)
ming024/FastSpeech2
Jun 2020 - Jul 2021
An implementation of Microsoft's "FastSpeech 2: Fast and High-Quality End-to-End Text to Speech"
Role in this project:
ML Engineer
ML EngineerContributions:35 commits, 1 PR, 34 pushes in 1 year
Contributions summary:Chung-ming primarily focused on modifying and improving the model architecture for the FastSpeech2 implementation. Their contributions included adding functionality for logging loss traces using TensorBoard. They also worked on modifying the data processing pipeline, particularly within the data loading and pre-processing steps for multiple datasets. Furthermore, the user integrated a HiFi-GAN vocoder for generating high-quality speech samples, enhancing the overall quality of generated audio.
text-to-speechend-to-endspeech-to-textspeech-recognitionhigh-quality
ming024/hierarchical_prosody_modeling
Aug 2020 - Jan 2021
Contributions:3 commits, 2 pushes in 4 months
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.