Soohwan Kim
ML Engineer - Operation
Seoul, South Korea
Summary
🤩
Rockstar🎓
Top SchoolSoohwan Kim is an AI engineer and founder with six years of hands-on experience building production-grade speech and natural language systems, currently working as an ML Engineer - Operation in Seoul. He co-founded TUNiB and led development of a proprietary Korean LLM ('Cheetah'), conversational AI products, and NLP APIs deployed to enterprise clients, while securing significant seed funding. Previously at Kakao Brain he contributed to Pororo and multilingual speech research, including Korean pre-training of Wav2Vec 2.0 and TTS/ASR systems. A pragmatic bridge between research and production, he independently managed model training, serving infrastructure, and operations at startup scale. His open-source contributions include core implementations and refactors for Korean ASR toolkits (KoSpeech, Conformer, OpenSpeech) and TTS components in Pororo, reflecting deep expertise in speech stack internals. Notably, he pairs competitive research achievements (EMNLP first-author paper and national challenge awards) with production delivery and maintainable code practices.
6 years of coding experience
2 years of employment as a software developer
Bachelor of Engineering - BE, Electronic and Communications Engineering (Major), Data Science (Minor), Bachelor of Engineering - BE, Electronic and Communications Engineering (Major), Data Science (Minor) at Kwangwoon University
Github Skills (34)
pytorch10
python10
machine-learning10
speech-synthesis10
mask-rcnn10
transformer-models10
deeplearning-ai10
deep-learning10
conformer10
natural-language-processing10
asr10
convolution10
faster-rcnn10
transformer10
speech-recognition10
Programming languages (7)
Github contributions (5)
sooftware/conformer
Jan 2021 - Sep 2022
[Unofficial] PyTorch implementation of "Conformer: Convolution-augmented Transformer for Speech Recognition" (INTERSPEECH 2020)
Role in this project:
ML Engineer
ML EngineerContributions:1 release, 77 commits, 36 PRs in 1 year 7 months
Contributions summary:Soohwan implemented a PyTorch-based Conformer model for speech recognition, as evidenced by the code changes across multiple files, including `modules.py`, `conv.py`, `model.py`, `feed_forward.py`, `encoder.py`, `attention.py`, and `embedding.py`. Their contributions included defining the model architecture, convolutional layers, attention mechanisms, feed-forward networks, and other necessary components. The user appears to be responsible for building the core structure of the Conformer model, which aligns with the repository's description of a PyTorch implementation for speech recognition.
pytorchconvend-to-endasrctc
sooftware/kospeech
Oct 2019 - Aug 2022
Open-Source Toolkit for End-to-End Korean Automatic Speech Recognition leveraging PyTorch and Hydra.
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:4 releases, 2161 commits, 74 PRs in 2 years 10 months
Contributions summary:Soohwan's contributions primarily involved refactoring and implementing code changes related to the core functionalities of the KoSpeech project. Specifically, the changes are centered around code naming conventions and refactoring, indicating a focus on improving code readability and maintainability. They also worked on testing and debugging the core algorithms related to the project, suggesting involvement in machine learning and speech recognition components.
asrctcspeech-recognitione2e-asrend-to-end
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.