Jong Wook Kim
OpenAI member of technical staff; co-author of CLIP and Whisper and maintainer of the Whisper repository
Jong Wook Kim is a member of technical staff at OpenAI, where Kim has worked since August 2019 on large-scale multimodal deep learning, including the CLIP, Jukebox and Whisper projects. Kim is the second author of the CLIP and Whisper papers, and as a maintainer of the openai/whisper repository announced the large-v3 and large-v3-turbo Whisper models.125678
Career
Kim worked as a software engineer at NCSOFT and then as a recommender system engineer and data scientist at Kakao, building distributed systems and Spark applications for recommendations. Kim later held research internships at Pandora in 2017 and Spotify in 2018.32
Kim joined OpenAI as a member of technical staff in August 2019. Kim's résumé describes developing and open-sourcing multimodal models such as Jukebox, CLIP and Whisper, building the Whisper API, and taking part in data engineering for GPT-4 with vision.32
Research
Kim's early research was in music information retrieval, including CREPE (2018), a convolutional model for pitch estimation. At OpenAI, Kim co-authored Jukebox (2020), a generative model for music.1039
Kim is second author of the CLIP paper (2021), which trains image and text encoders together from natural language supervision, and of the Whisper paper (2022), on speech recognition trained with large-scale weak supervision. In the Whisper repository, Kim announced large-v3 in 2023, which uses 128 Mel frequency bins, and large-v3-turbo in 2024, which reduces the decoder to 4 layers.5678
Work
CLIP ViT-L/14Open weights
Second author, after Alec Radford, of the 2021 paper "Learning Transferable Visual Models From Natural Language Supervision", which introduced CLIP.53
Whisper large-v3Open weights
Second author of the Whisper paper, and posted the November 2023 announcement of the large-v3 model in the openai/whisper repository as a maintainer.67
Whisper large-v3-turboOpen weights
Posted the October 2024 announcement of large-v3-turbo, a version of large-v3 with 4 decoder layers instead of 32, in the openai/whisper repository as a maintainer.8
Career timeline
May 2011
Began work as a software engineer at NCSOFT Corporation.3
Jun 2014
Joined Kakao as a recommender system engineer and data scientist.3
Feb 2018
Posted CREPE, a convolutional representation for pitch estimation, as first author.10
Aug 2019
Apr 2020
Co-authored Jukebox, a generative model for music, at OpenAI.9
Feb 2021
Posted the CLIP paper as second author with Alec Radford and colleagues.5
Dec 2022
Posted the Whisper paper on large-scale weakly supervised speech recognition as second author.6
Nov 6, 2023
Announced the Whisper large-v3 model in the openai/whisper repository.7
Oct 1, 2024
Announced the Whisper large-v3-turbo model in the openai/whisper repository.8
Related people
More sourced profiles at the same organization, in the same field, or from the same school.
- Aditya RameshAlso at OpenAI
- Alec RadfordAlso at OpenAI
- Alex NicholAlso at OpenAI
- Alexandre DéfossezSpeech and audio
- Eric MitchellAlso at OpenAI
- Alexei BaevskiSpeech and audio
- Greg BrockmanAlso at OpenAI
- David HarwathSpeech and audio
- Jakub PachockiAlso at OpenAI
- Wei-Ning HsuSpeech and audio
- Łukasz KaiserAlso at OpenAI
- Xu TanSpeech and audio
Is this you?
Profiles cover public professional work only, and every line cites a source. If something is out of date, missing or wrong, send a correction with a link to an official page and we will update it. Suggest a correction
Add this badge to your website or GitHub profile README. It links back to this page. It is not an endorsement or a rating.
Markdown (for a README)
[](https://unitedstatesofamericasuperintelligence.com/people/jong-wook-kim/)
HTML
<a href="https://unitedstatesofamericasuperintelligence.com/people/jong-wook-kim/"><img src="https://unitedstatesofamericasuperintelligence.com/badges/profile.svg" alt="Jong Wook Kim — Profile on USASI" height="20"></a>
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.