□ DGIST (President Kunwoo Lee) announced that it has been selected for a new project under the Ministry of Science and ICT’s “Next-Generation AI+S&T (Science and Technology) Foundational Technology Development” program. Through the project, DGIST will undertake the full-scale development of foundational technologies for an “AI Mechanician”—an AI system capable of autonomously identifying and interpreting the behavior of fluids, heat, matter, and other phenomena within complex porous structures.
□ The project will receive a total of KRW 5 billion in government R&D funding over five years, from 2026 to 2031. Professor Jaesok Yu of the DGIST Department of Robotics and Mechatronics Engineering will serve as principal investigator, with Professors Inkyu Moon, Hoe Joon Kim, and Sohyun Jung of the same department providing as co-researchers. Professor Sang Hyun Park’s team at Pohang University of Science and Technology (POSTECH) will participate as a joint R&D institution.
□ Porous structures are ubiquitous across science and engineering, including in battery electrodes and separators, catalyst layers, sound-absorbing materials, and biological tissues. These structures are difficult to analyze because the sizes and arrangements of their pores are intricately intertwined and vary from one sample to another. Conventional high-fidelity numerical analysis methods require enormous amounts of time to solve even a single geometric configuration, while the repeated calculations needed for design optimization have emerged as a major bottleneck throughout the R&D process.
□ The “AI Mechanician” being developed by the research team is an innovative system designed to emulate the reasoning process of an experienced mechanician. At its core is an integrated Agentic AI architecture that combines modules that extract only the necessary structural features from visualization images, autonomously identify the dominant physical phenomena and infer key physical properties, and independently verify conservation laws and consistency.
□ In particular, the team plans to overcome the limitations of existing AI models that are “fast but difficult to trust” by designing the model to present information on physical validity and uncertainty alongside its prediction results. The research team has strong capabilities to independently complete the entire research cycle, from specimen design and fabrication to experiments, data generation, and AI training and validation, without relying on external resources. Going forward, the team also plans to establish open benchmarks and evaluation code that enable fair comparisons among reference numerical analyses, measured data, and AI predictions, thereby providing a foundation for follow-up research.
□ “While existing AI-based analysis models have focused on rapidly solving predefined problems, the AI Mechanician aims to autonomously determine which physical phenomena are dominant and even verify whether its answers are valid,” said Professor Jaesok Yu, the principal investigator. “By enabling design exploration that has been difficult to attempt because of computational costs, we will dramatically accelerate R&D across various fields including batteries, energy, and healthcare.”
□ “When mechanicians look at a structure, they use their experience to determine the level of analysis required, but until now, that judgment has existed only in their minds,” said Professor Sohyun Jung, a co-researcher on the project. “Since this research seeks to translate that decision-making process itself into data and models, we expect not only a reduction in analysis time but also a significant expansion in the range of designs that researchers can try.”
□ Meanwhile, DGIST has designated “Physical AI,” which combines the laws of physics with AI, as a future strategic field and is focusing on securing foundational technologies in this area. In particular, the DGIST College of AI will establish three departments—Physical AI, Medical AI, and AI-X—to cultivate interdisciplinary talent, with admissions beginning in the 2027 academic year. The foundational AI analysis technologies to be developed through this project are expected to be rapidly incorporated into education and research. DGIST plans to continue strengthening its leading position in key technologies that will transform the paradigm of scientific and engineering analysis, thereby contributing to strengthening Korea’s competitiveness in national strategic technologies.
