Gwanghoon Yoo유광훈

I started by building AI that understands human language. Today, I study how people can learn and grow with AI. My work spans natural language understanding and sentiment analysis, reasoning systems that combine language models with symbolic knowledge, Graph RAG for retrieving and composing knowledge, and agent architectures that improve their work through execution evidence.

My central interest is what knowledge AI uses, what evidence supports its decisions, and how execution results inform the next task. I have designed architectures that turn agent experience into reusable knowledge, compose relevant knowledge into execution plans, and use evaluation results to improve subsequent behavior. My work brings knowledge, reasoning, and evaluation together in the design of the whole system. I also define technical roadmaps and coordinate development across machine learning, backend, product, and research teams.

Today, I study how people can learn and grow through their experience of using AI. As AI becomes more capable, people should also be able to understand its results, intervene when needed, and take responsibility for their decisions. I research ways to adapt explanations to a person’s understanding, as well as representations that serve both explanations for people and memory for agents.

I live by questions, not answers. FACE, Freedom, Authenticity, Creativity, and Expandability, guides how I work. I choose problems with purpose, commit to them with sincerity, and explore different approaches. The freedom I seek includes using the knowledge and experience I develop to help others succeed.