Career Profile
My primary research area focuses on anomaly detection through transfer learning and knowledge distillation in situations where datasets are scarce in computer vision. I am researching methods to create uncertainty maps based on the inherent uncertainty in the network, derived from learned normal road situations, and optimize these to extract semantically useful pixel information for driver safety. I am interested in optimizing semantic segmentation networks that emerge from probabilistic models for abnormal situations captured by front-facing cameras in vehicles. Additionally, I have a significant interest in metaverse content creation based on various generative models. In particular, I have been highly interested in data-driven generative models that model the entire playable game framework, including user interactions, beyond asset creation.
Education
During my doctoral research, I focused on the development and implementation of Explainable Artificial Intelligence (XAI) technologies across various sectors, including energy, education, and security. My projects ranged from integrating AI for anomaly detection and explanation in South Korea’s power systems, which facilitated the identification of abnormalities and enhanced system reliability, to leading the creation of a smart AI-based learning platform for English language acquisition, and developing virtual training environments for counter-terrorism with the Korean National Police Agency. These diverse applications not only highlighted the versatility of XAI in different operational contexts but also advanced the transparency and efficiency of AI systems in critical and everyday scenarios.
I conducted research to develop a deep learning-based language model for open-domain Korean natural language sentences. Utilizing this model, I was responsible for generating vectors for Korean sentences and calculating their similarities, thus providing an auxiliary decision-making tool for AI models. This effort was part of a larger project aimed at strengthening Korea’s AI ecosystem.
During my undergraduate studies, I demonstrated outstanding achievements across various major subjects, equipping myself with a comprehensive academic understanding and proficiency. Particularly, with a deep knowledge in computer science, I developed a strong interest and passion for graphics and computer game development. This interest led to the topic of my graduation thesis, where I presented an engaging game format by integrating cutting-edge graphic technology with original game design principles. Throughout this process, I extensively researched not only programming skills but also methods to optimize visual representation and user experience. The knowledge and experience gained during my undergraduate course provided me with the technical foundation and creative approach necessary to solve complex problems in software development.
Experiences
I conducted research to develop a deep learning-based language model for open-domain Korean natural language sentences. Utilizing this model, I was responsible for generating vectors for Korean sentences and calculating their similarities, thus providing an auxiliary decision-making tool for AI models. This effort was part of a larger project aimed at strengthening Korea’s AI ecosystem.
- Link to project: etri.re.kr
The project was undertaken by the Korea Electric Power Corporation’s Power Research Institute, focusing on researching AI technologies related to energy. My role involved anomaly detection and explanation generation, proposing technology for identifying and generating explanations for abnormal situations in power system field situation images.
- Link to project: etnews.com
The project aims to develop the “AILA” platform for a non-contact English speaking learning environment. My role encompassed integrating and managing the Unity and chatbot development environments, along with providing a personalized experience through the fusion of 3D avatar chatbot technology.
- Link to project: todayan.com
In a research program supported by the Korean National Police Agency, I am involved in creating an application framework for generating virtual environments based on manuals and developing similarity-based manual analysis and simulation AI algorithms.
- Link to project: dt.co.kr
The purpose of this project is to address the “black box” nature of deep learning methods in fields requiring reliability and transparency, such as the military, medical, and safety sectors. My work focused on image anomaly detection for industrial images based on generative models.
- Link to project: ku.ac.ae