Haohe Liu
Research Scientist at Meta
Redmond, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolHaohe Liu is a research scientist with seven years of experience specializing in generative audio and speech restoration, currently based in Redmond and working at Meta. He combines academic rigor from a PhD program at the University of Surrey with hands-on industry experience including internships at Microsoft and ByteDance and a recent Meta research stint in Paris. Haohe is an active open-source contributor—leading engineering work on projects like AudioLDM (text-to-audio generation) and VoiceFixer (speech restoration)—where he implemented new generator architectures, sampling pipelines, and audio-processing improvements. His skill set bridges deep learning research and production-ready model engineering, with a track record of shipping practical improvements to vocoders and generation pipelines. Colleagues can expect a scientist who moves quickly from novel models to usable code, often surfacing non-obvious trade-offs between synthesis quality and pipeline efficiency.
7 years of coding experience
Bachelor of Engineering - BE Computer Science and Engineering, Bachelor of Engineering - BE Computer Science and Engineering at Northwestern Polytechnical UniversityDoctor of Philosophy - PhD, Doctor of Philosophy - PhD at University of Surrey
Github Skills (15)
pytorch10
machine-learning10
speech-enhancement10
deep-learning10
speech-processing10
python10
n9
rnn-model9
audio-processing9
mask-rcnn9
faster-rcnn9
gtts8
freetts8
super-resolution7
git3
Programming languages (6)
Github contributions (5)
haoheliu/voicefixer
Sep 2021 - Sep 2022
General Speech Restoration
Role in this project:
ML Engineer
ML EngineerContributions:3 releases, 8 reviews, 64 commits in 1 year
Contributions summary:Haohe primarily contributed to the core functionality of the speech restoration project by modifying and updating the neural network models. They added a new generator model, updated the model architecture, and made improvements to the restoration strategies. The user also addressed bug fixes, updated dependencies, and added features such as support for other vocoders.
restorationdereverberationspeech-analysisspeech-to-textspeech-recognition
haoheliu/AudioLDM
Jan 2023 - Mar 2023
AudioLDM: Generate speech, sound effects, music and beyond, with text.
Role in this project:ML Engineer
ML EngineerContributions:6 reviews, 76 commits, 26 PRs in 1 month
Contributions summary:Haohe's commits primarily involve modifications to the `audioldm` project, which focuses on audio generation using text prompts. The user implemented changes to various python scripts including text2sound.py for generating audio, the main ldm.py, and the pipeline.py file demonstrating modifications related to the model, its sampling, and the generation process. Additionally, the user made modifications to the audio processing related functions within the codebase.
audio-generation
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.