Wonkyum Lee
Member Of Technical Staff at xAI
San Francisco Bay Area United States
Summary
👤
Senior🎓
Top SchoolWonkyum Lee is a machine learning and speech processing specialist with over a decade of experience building production-grade ASR, TTS, and NLP models and leading ML teams from research to deployment. He co-founded an agentic AI startup focused on scaling senior care and now contributes at xAI, blending product-first engineering with cutting-edge research. At Gridspace he led development of Grace, a voice AI agent deployed for Fortune 500 contact centers, and has hands-on low-level expertise—contributing CUDA kernels to the widely used Kaldi ASR codebase. Trained at Carnegie Mellon and KAIST, he bridges deep academic research in noisy and distant speech with pragmatically scalable systems engineering. Known for shipping safety- and outcome-oriented voice AI that ties technical rigor to measurable customer and health impacts.
10 years of coding experience
16 years of employment as a software developer
Bachelor of Engineering (B.E.) Radio and Communications Engineering, Bachelor of Engineering (B.E.) Radio and Communications Engineering at Korea UniversityMaster of Science (MSc) Electrical and Computer Engineering, Master of Science (MSc) Electrical and Computer Engineering at Carnegie Mellon UniversityMS Signal Processing (Detection and Estimation) Electrical Engineering, MS Signal Processing (Detection and Estimation) Electrical Engineering at Korea Advanced Institute of Science and Technology
English, Korean
Github Skills (7)
cuda10
mat10
machine-learning10
speech-recognition10
c-language10
matrix10
cprogramming-language10
Programming languages (4)
Github contributions (5)
kaldi-asr/kaldi
May 2016 - Jan 2020
kaldi-asr/kaldi is the official location of the Kaldi project.
Role in this project:
Back-end Developer / ML Engineer
Back-end Developer / ML EngineerContributions:14 PRs, 35 comments, 4 issues in 3 years 8 months
Contributions summary:Wonkyum contributed to the CUDA kernels within the Kaldi project, indicating involvement in the low-level implementation of speech recognition algorithms. They modified the `cu-kernels.cu`, `cu-kernels.h`, and `cu-matrix.cc` files, adding and updating CUDA kernels for matrix operations. The commits include changes to existing functions and the addition of new functions, specifically for matrix operations and operations used in neural network layers such as softmax.
cudakaldiasrspeech-to-textkaldi-asr
cricketclub/kaldi
Apr 2017 - Jul 2020
This is now the official location of the Kaldi project.
Contributions:20 pushes, 9 branches, 2 comments in 3 years 3 months
kaldilocation
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.