Junhyuk Oh
I'm a co-founder of Ineffable Intelligence. I'm interested in building agents that learn from experience and scalable algorithms through which intelligence can emerge.
Before Ineffable Intelligence, I was a senior staff research scientist at Google DeepMind. I received my Ph.D. from the University of Michigan, where I was advised by Honglak Lee and Satinder Singh.
I go by Jun pronounced like the month “June”

News
Co-founded Ineffable Intelligence.
DiscoRL was published in Nature.
Invited talks
RLC 2026 Workshop on Automated RLMontréal, Canada
ICML Workshop on Decision Awareness in Reinforcement LearningBaltimore, USA
Selected work
- 2025Discovering state-of-the-art reinforcement learning algorithmsNature
- 2023Gemini: A Family of Highly Capable Multimodal ModelsTechnical report
- 2020Discovering Reinforcement Learning AlgorithmsNeurIPS
- 2019Grandmaster level in StarCraft II using multi-agent reinforcement learningNature
- 2017Value Prediction NetworkNeurIPS