Hyeonwoo Noh
Hyeonwoo Noh is a researcher in visual understanding who moved from a POSTECH doctorate into OpenAI's work on GPT-4. His early work includes a 2015 paper on a deconvolution network for semantic segmentation, and he completed his PhD at POSTECH in 2019. He joined OpenAI as a research scientist in September 2019, and in 2023 he led visual understanding research for GPT-4. In 2025 he founded and led the Computer-Using Agent team behind Operator and ChatGPT Agent. As of October 2026, he is a former research lead at OpenAI, and that appointment ended in January 2026.
- Works
- 15
- Citations
- 5,951
- h-index
- 12
Publication numbers from OpenAlex, as of 15 October 2026.
Timeline
-
2007–2013
Computer science at POSTECH
He studies computer science and engineering at POSTECH from March 2007 to July 2013.
-
2015–2019
Doctoral study at POSTECH
He completes a PhD in computer science and engineering at POSTECH from March 2015 to July 2019.
-
2015
A deconvolution network for segmentation
In 2015 he, Seunghoon Hong and Bohyung Han publish Learning Deconvolution Network for Semantic Segmentation, which appears at ICCV 2015.
-
2017
DELF for image retrieval
Together with Andre Araujo, Jack Sim and Tobias Weyand, he co-writes Large-Scale Image Retrieval with Attentive Deep Local Features, which appears at ICCV 2017.
-
September 2019
A research scientist at OpenAI
He joins OpenAI as a research scientist in September 2019, and the appointment ends in January 2026.
-
2023
Visual understanding for GPT-4
In 2023 he leads visual understanding research for GPT-4, and the GPT-4 Technical Report lists him as an author.
-
2025
The Computer-Using Agent team
In 2025 he works on Operator and ChatGPT Agent, and he founds and leads the Computer-Using Agent team behind them.
-
As of October 2026
Former research lead at OpenAI
He is a former research lead at OpenAI.
Sources and credits
Photo: We have no published photo of Hyeonwoo Noh yet. Send us a free-licensed photo.