Yun Tang
Researcher at Hippocratic AI
San Francisco Bay Area United States
Summary
🤩
Rockstar🎓
Top SchoolYun Tang is a researcher and machine learning engineer with 10 years of experience building speech and language systems, currently working in the San Francisco Bay Area. He combines deep academic training (Ph.D. in Electrical Engineering) with industry research roles at companies including Meta, Samsung Research America, Apple, and Nuance, focusing on pattern recognition, automatic speech recognition, machine translation, and NLP. Yun has practical systems expertise in C/C++, Python, and scripting languages, and has shipped production-oriented research such as speech-to-text and speech-translation components. As an active open-source contributor, he extended Facebook Research’s widely used fairseq toolkit with joint speech/text training, raw audio support, and decoding/alignment fixes—improvements that bridge research models and usable pipelines. Colleagues know him for translating novel algorithms into robust code and for attention to integration details like source tags and EOS-handling that often make research models work in practice.
10 years of coding experience
16 years of employment as a software developer
Ph. D., EE, Ph. D., EE at Chinese Academy of SciencesB. S., B. S. at Zhejiang University
Github Skills (9)
transformers10
sequence-to-sequence10
pytorch10
machine-learning10
speech-recognition10
artificial-intelligence10
python10
fairseq10
nlp9
Programming languages (3)
Github contributions (5)
facebookresearch/fairseq
Jul 2021 - Jun 2022
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
Role in this project:
ML Engineer
ML EngineerContributions:11 reviews, 8 commits, 22 comments in 10 months
Contributions summary:Yun contributed to the development of a speech-to-text system within the fairseq framework. Their work involved implementing speech/text joint training capabilities, including adding features for raw audio generation, fixing bugs related to speech-to-text decoding, and integrating scripts for joint training tasks. Furthermore, they updated the existing example code to align with the latest fairseq codebase, and provided support for alignment and the use of source language IDs. This included adding support for skipping the EOS token and adding functionality around source tags.
pytorchnlpsequencepythontransformer-architecture
yuntang/NeMo
Nov 2023 - Nov 2023
Contributions:1 push in 1 day
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.