Keunwoo Choi
New York, New York, United States
Summary
🤩
RockstarKeunwoo Choi is an AI engineer and researcher with a decade of experience building production-grade machine learning systems at the intersection of audio, music, and large language models. He was a founding co-lead of Genentech’s Prescient-LM, owning end-to-end LLM strategy, training, evaluation, deployment, and user studies, and has continued to bridge research and product in industry roles at ByteDance, Spotify, and startups. A prolific open-source contributor, his work on kapre, librosa, and Keras added audio-aware layers and features (STFT, melspectrogram, SpecAugment, spectral_flatness) that are widely used by the ML-audio community. He teaches and advises—serving as adjunct faculty at KAIST and NYU and advising Gaudio Lab—bringing academic rigor to applied AI. Based in New York, he also runs a music-focused project, seoulunderground.live, reflecting a persistent creative curiosity about music culture and tech. Known for combining signal-processing depth with scalable ML engineering, he excels at turning domain expertise in audio into practical models and tools.
10 years of coding experience
Github Skills (26)
scipy10
convolutional-neural-networks10
spectrogram10
python10
feature-extraction10
signal-processing10
machine-learning10
rnn-model10
recurrent-neural-networks10
audio-processing10
n10
numpy10
digital-signal-processing10
mask-rcnn10
audio10
Programming languages (14)
Github contributions (5)
keunwoochoi/kapre
Dec 2016 - Jan 2022
kapre: Keras Audio Preprocessors
Role in this project:
Back-end Developer
Back-end DeveloperContributions:9 releases, 19 reviews, 276 commits in 5 years 2 months
Contributions summary:Keunwoo primarily worked on adding and modifying time-frequency representations and data augmentation layers. They developed STFT and Melspectrogram layers and introduced the SpecAugment layer for audio data augmentation, indicating a focus on audio processing and deep learning. Their contributions also included improvements to the documentation and examples, indicating their interest in project maintainability.
audiokerasspectrogrammelspectrogrampreprocess
keunwoochoi/music-auto_tagging-keras
Aug 2016 - Jul 2018
Music auto-tagging models and trained weights in keras/theano
Role in this project:ML Engineer
ML EngineerContributions:92 commits, 1 PR, 77 pushes in 1 year 11 months
Contributions summary:Keunwoo primarily contributed to the development of music auto-tagging models using Keras. Their commits include the creation of convolutional neural network (CNN) and recurrent neural network (RNN) models, along with the implementation of example code for tag prediction. The user also added functionality to load and process audio data for model training and inference, showcasing a focus on the end-to-end machine learning pipeline. They also refactored and updated the code base to be compatible with the newer Keras applications.
kerastheano
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.