%41%49
-
KAIST Develops AI Technology That Predicts and Assembles Cell Drug Responses Like Lego Blocks
<(From left) Dr. Younghyun Han, (top center) Dr. Chun-Kyung Lee, (bottom center) Prof. Kwang-Hyun Cho,Ph.D. candidate Hyunjin Kim>
Controlling the state of a cell in a desired direction is one of the central challenges in life sciences, including drug development, cancer treatment, and regenerative medicine. However, identifying the right drug or genetic target for that purpose is extremely difficult. To address this, researchers at KAIST have mathematically modeled the interaction between cells and drugs in a modular “Lego block” manner—breaking them down and recombining them—to develop a new AI technology that can predict not only new cell–drug reactions never before tested but also the effects of arbitrary genetic perturbations.
KAIST (President Kwang Hyung Lee) announced on the 16th of October that a research team led by Professor Kwang-Hyun Cho of the Department of Bio and Brain Engineering has developed a generative AI-based technology capable of identifying drugs and genetic targets that can guide cells toward a desired state.
“Latent space” is an invisible mathematical map used by image-generating AI to organize the essential features of objects or cells. The research team succeeded in separating the representations of cell states and drug effects within this space and then recombining them to predict the reactions of previously untested cell–drug combinations. They further extended this principle to show that the model can also predict how a cell’s state would change when a specific gene is regulated.
The team validated this approach using real experimental data. As a result, the AI identified molecular targets capable of reverting colorectal cancer cells toward a normal-like state, which the team later confirmed through cell experiments.
This finding demonstrates that the method is not limited to cancer treatment—it serves as a general platform capable of predicting various untrained cell-state transitions and drug responses. In other words, the technology not only determines whether or not a drug works but also reveals how it functions inside the cell, making the achievement particularly meaningful.
<Latent Space Direction Vector–Based Cell Transition Modeling>
The research provides a powerful tool for designing methods to induce desired cell-state changes. It is expected to have broad applications in drug discovery, cancer therapy, and regenerative medicine, such as restoring damaged cells to a healthy state.
Professor Kwang-Hyun Cho stated, “Inspired by image-generation AI, we applied the concept of a ‘direction vector,’ an idea that allows us to transform cells in a desired direction.” He added, “This technology enables quantitative analysis of how specific drugs or genes affect cells and even predicts previously unknown reactions, making it a highly generalizable AI framework.”
The study was conducted with Dr. Younghyun Han, Ph.D. candidate Hyunjin Kim, and Dr. Chun-Kyung Lee of KAIST. The research findings were published online in Cell Systems, a journal by Cell Press, on October 15.
※ Paper title: “Identifying an Optimal Perturbation to Induce a Desired Cell State by Generative Deep Learning” (DOI: 10.1016/j.cels.2025.101405)
The study was supported by the National Research Foundation of Korea (NRF) through the Ministry of Science and ICT’s Mid-Career Researcher Program and the Basic Research Laboratory (BRL) Program.
2025.10.16 View 3086 -
Federated Learning AI Developed for Hospitals and Banks Without Personal Information Sharing
< (From bottom left) KAIST Ph.D. Candidate Yoonho Lee, Integrated M.S./Ph.D. Candidate Sein Kim, Ph.D. Candidate Sungwon Kim, Ph.D. Candidate Junseok Lee, Ph.D. Candidate Yunhak Oh, (From top right) Ph.D. Candidate Namkyeong Lee, UNC Chapel Hill Ph.D. Candidate Sukwon Yun, Emory University Professor Carl Yang, KAIST Professor Chanyoung Park >
Federated Learning was devised to solve the problem of difficulty in aggregating personal data, such as patient medical records or financial data, in one place. However, during the process where each institution optimizes the collaboratively trained AI to suit its own environment, a limitation arose: the AI became overly adapted to the specific institution's data, making it vulnerable to new data. Our university research team has presented a solution to this problem and confirmed its stable performance not only in security-critical fields like hospitals and banks but also in rapidly changing environments such as social media and online shopping.
KAIST announced on October 15th that the research team led by Professor Chanyoung Park of the Department of Industrial and Systems Engineering has developed a new learning method that fundamentally solves the chronic performance degradation problem of Federated Learning, significantly enhancing the Generalization performance of AI models.
Federated Learning is a method that allows multiple institutions to jointly train an AI without directly exchanging data. However, a problem occurs when each institution fine-tunes the resulting joint AI model to its local setting. This is because the broad knowledge acquired earlier is diluted, leading to a Local Overfitting problem where the AI becomes excessively adapted only to the data characteristics of a specific institution.
For example, if several banks jointly build a 'Collaborative Loan Review AI,' and one specific bank performs fine-tuning focusing on corporate customer data, that bank's AI becomes strong in corporate reviews but suffers from local overfitting, leading to degraded performance in reviewing individual or startup customers.
Professor Park's team introduced the Synthetic Data method to solve this. They extracted only the core and representative features from each institution's data to generate virtual data that does not contain personal information and applied this during the fine-tuning process. As a result, each institution's AI can strengthen its expertise according to its own data without sharing personal information, while maintaining the broad perspective (generalization performance) gained through collaborative learning.
<Figure 1. Federated Learning is a distributed learning method where multiple institutions collaboratively train a joint Artificial Intelligence model without directly sharing their data. Each institution trains its individual AI model using its local data (Institution 1, 2, 3 Data). Afterward, only the trained model information, not the original data, is securely aggregated to a central server to construct a high-performing 'Joint AI Model.' This method allows for the effect of training with diverse data while protecting the privacy of sensitive information>
< Figure 2. The Local Overfitting problem occurs during the process of fine-tuning the 'Joint AI Model' built through Federated Learning with each institution's data. For example, Institution 3 can fine-tune the joint AI with its own data (Type 0, 2) to create an expert AI for those types, but in the process, it forgets the knowledge about data (Type 1) that other institutions had (Information Loss). In this way, each institution's AI becomes optimized only for its own data, gradually losing the ability (generalization performance) to solve other types of problems that were obtained through collaboration. >
The research results showed that this method is particularly effective in fields where data security is crucial, such as healthcare and finance, and also demonstrated stable performance in environments where new users and products are continuously added, like social media and e-commerce. It proved that the AI could maintain stable performance without confusion even if a new institution joins the collaboration or data characteristics change rapidly.
< Figure 3. The technology proposed by the research team solves the local overfitting problem by utilizing Synthetic Data. When each institution fine-tunes its AI with its own data, it simultaneously trains with 'Global Synthetic Data' created from the data of other institutions. This synthetic data acts as a kind of 'Vaccine' to prevent the AI from forgetting information not present in the local data (e.g., Type 2 in the image), helping the AI to gain expertise on specific data while retaining a broad view (generalization performance) to handle other types of data. >
Professor Chanyoung Park of the Department of Industrial and Systems Engineering said, "This research opens a new path to simultaneously ensure both expertise and versatility for each institution's AI while protecting data privacy," and "It will be a great help in fields where data collaboration is essential but security is important, such as medical AI and financial fraud detection AI."
The research was primarily authored by Graduate School of Data Science student Sungwon Kim and co-authored by Professor Chanyoung Park as the corresponding author. It was recognized for its excellence by being selected for an Oral Presentation, which is reserved for the top 1.8% of outstanding papers, at the International Conference on Learning Representations (ICLR) 2025, a top-tier academic conference in the field of Artificial Intelligence held in Singapore last April.
※ Paper Title: Subgraph Federated Learning for Local Generalization, https://doi.org/10.48550/arXiv.2503.03995
Meanwhile, this research is a result of projects supported by the Institute of Information & Communications Technology Planning & Evaluation (IITP) — the 'Robust, Fair, and Scalable Data-Centric Continual Learning' project, the National Research Foundation of Korea (NRF) — the 'Graph Foundation Model: Graph-based Machine Learning Applicable to Various Modalities and Domains' project, and the 'Data Science Convergence Talent Fostering Program.'
2025.10.15 View 2069 -
KAIST Develops Multimodal AI That Understands Text and Images Like Humans
<(From Left) M.S candidate Soyoung Choi, Ph.D candidate Seong-Hyeon Hwang, Professor Steven Euijong Whang>
Just as human eyes tend to focus on pictures before reading accompanying text, multimodal artificial intelligence (AI)—which processes multiple types of sensory data at once—also tends to depend more heavily on certain types of data. KAIST researchers have now developed a new multimodal AI training technology that enables models to recognize both text and images evenly, enabling far more accurate predictions.
KAIST (President Kwang Hyung Lee) announced on the 14th that a research team led by Professor Steven Euijong Whang from the School of Electrical Engineering has developed a novel data augmentation method that enables multimodal AI systems—those that must process multiple data types simultaneously—to make balanced use of all input data.
Multimodal AI combines various forms of information, such as text and video, to make judgments. However, AI models often show a tendency to rely excessively on one particular type of data, resulting in degraded prediction performance.
To solve this problem, the research team deliberately trained AI models using mismatched or incongruent data pairs. By doing so, the model learned to rely on all modalities—text, images, and even audio—in a balanced way, regardless of context.
The team further improved performance stability by incorporating a training strategy that compensates for low-quality data while emphasizing more challenging examples. The method is not tied to any specific model architecture and can be easily applied to various data types, making it highly scalable and practical.
<Model Prediction Changes with a Data-Centric Multimodal AI Training Framework>
Professor Steven Euijong Whang explained, “Improving AI performance is not just about changing model architectures or algorithms—it’s much more important how we design and use the data for training.” He continued, “This research demonstrates that designing and refining the data itself can be an effective approach to help multimodal AI utilize information more evenly, without becoming biased toward a specific modality such as images or text.”
The study was co-led by doctoral student Seong-Hyeon Hwang and master’s student Soyoung Choi, with Professor Steven Euijong Whang serving as the corresponding author. The results will be presented at NeurIPS 2025 (Conference on Neural Information Processing Systems), the world’s premier conference in the field of AI, which will be held this December in San Diego, USA, and Mexico City, Mexico.
※ Paper title: “MIDAS: Misalignment-based Data Augmentation Strategy for Imbalanced Multimodal Learning,” Original paper: https://arxiv.org/pdf/2509.25831
The research was supported by the Institute for Information & Communications Technology Planning & Evaluation (IITP) under the projects “Robust, Fair, and Scalable Data-Centric Continual Learning” (RS-2022-II220157) and “AI Technology for Non-Invasive Near-Infrared-Based Diagnosis and Treatment of Brain Disorders” (RS-2024-00444862).
2025.10.14 View 3738 -
Sharing Failures Makes Challenges Easier, Proposal for a National Campaign for Global Failure Day
KAIST announced that it will launch a national campaign on 'Global Failure Day,' October 13th, to encourage anyone in the nation to share their small, everyday failures.
KAIST President Kwang Hyung Lee emphasized, "A culture unafraid of failure is the foundation of innovation. I hope that for just one day, October 13th, people recall and share the small failures they experienced today. That moment can be the starting point for a new challenge."
'Global Failure Day' is a commemorative day that began in 2010 by students at Aalto University in Finland under the spirit of "Failure is inherent in the nature of challenge; let's respect failure." At the time, amid the collapse of Nokia and job insecurity, it garnered significant social support and spread as a national campaign. It continued in countries like Germany, the UK, and Canada, and has now established itself globally as a day for reflecting on failure.
Since the establishment of the KAIST Failure Lab, there has been a clear shift in the perception of failure. According to a survey conducted by the Failure Lab in December last year, 73.9% of KAIST members responded that the atmosphere encourages new challenges, which is twice the Korean social average (35.6%). Furthermore, 52% responded that it is a "place tolerant of failure," much higher than the Korean average (20.5%).
To spread this shift in perception nationwide, President Lee personally posted a message on social media on the 10th, sharing his own story of a failure where he felt embarrassed after having a donation rejected, and proposed participation. Additionally, the KAIST Failure Lab announced a 'Failure Sharing Action Proposal' that anyone can easily participate in on a daily basis.
The main proposals include: △ Sharing 'Today's Failure' with family and friends; △ Having a 'One-Line Failure Sharing' time at work or in a gathering; △ Posting small failure stories on social media; △ Sharing photos/videos of disastrous cooking or silly mistakes; and △ Creating memes that humorously express failure.
Seongho Cho, Director of the Failure Lab, explained, "Simply sharing failures lightly can change the attitude towards them. The fact that the failure acceptance rate among KAIST members is twice as high as the general public is also thanks to this culture."
Since its establishment in 2021, the KAIST Failure Lab has promoted a culture of failure sharing within the university through various programs, such as the 'Failed Project Bragging Contest,' failure essay contests, and 'Failure Photo Voice.' It has been conducting perception surveys among KAIST members every two years since 2022, and in last year's survey, over 80% of respondents said the lab's activities contributed to improving perception, resilience, and flexibility.
Based on these achievements, the scope of the activities is being expanded nationwide this year. Notably, the top 10 teams in the 'AI × Failure Idea Contest' for university/graduate students are scheduled to present their ideas at the 'Failure Conference' to be held at KAIST on November 7th.
President Lee stated, "KAIST will continue to broaden the culture of reflecting on and sharing failure together with the public."
More details can be found on the KAIST Failure Lab website (https://caf.kaist.ac.kr).
2025.10.13 View 1596 -
AI Nüshu Wins International Award
< (From left) Dr. Yuqian Sun, Professor Chang-Hee Lee of the Department of Industrial Design, and Ali Asadipour, Director of CSRC at the Royal College of Art >
'Nüshu (女書)' is the world's only women's script, a unique writing system created autonomously by women in Hunan Province, China, starting around the 19th century. These women, excluded from Hanzi education, used it to record their lives and communicate with each other. A research team from KAIST participated in the 'AI Nüshu (女书)' project, which combines the script's significance (creation amidst oppression, female solidarity, linguistic experimentation) with modern technology, winning a prestigious international award often called the 'Academy Award of the media art world.'
KAIST announced on the 10th that the 'AI Nüshu' project, jointly conducted by Professor Chang-Hee Lee's research team from the Department of Industrial Design and Ali Asadipour, Director of the Computer Science Research Center at the Royal College of Art (RCA), was selected for the Honorary Mention in the Digital Humanity category at the 'Prix Ars Electronica 2025,' the world's highest-level media art festival.
< Installation image of 'AI Nüshu' >
The 'Prix Ars Electronica,' known as the 'Academy Award of the media art world,' is the premier international media art competition held annually in Linz, Austria. This competition, which discovers innovative works spanning the boundaries of art and science, saw 3,987 submissions from 98 countries this year, with only two works receiving the honor in the Digital Humanity category.
The award-winning work, 'AI Nüshu (女书),' is based on 'Nüshu,' the world's only women's script created by Chinese women who were excluded from literacy education to record and communicate their lives.
The KAIST research team and collaborators combined this script with Computational Linguistics to create an installation that visitors can directly experience.
The artificial intelligence within the artwork learns the communication methods of pre-modern Chinese women and generates its own new language. This is regarded as a symbol of resistance against the patriarchal order and a feminist endeavor that moves beyond Western-centric views on language.
< Example of the same sentence expressed in English, Chinese, Nüshu, and AI Nüshu >
It also received high praise for artistically presenting the possibility of machines creating new languages, going beyond the preconception that 'only humans create language.'
Dr. Yuqian Sun of the Royal College of Art expressed her feelings, saying, "Although there were many difficulties in my life and research process, I feel great reward and emotion through this award."
Professor Chang-Hee Lee of the KAIST Department of Industrial Design stated, "It is very meaningful that this contemplative art, born from the intersection of history, humanities, art, and technology, has led to such a globally prestigious award."
Detailed information about the project can be found on the official Prix Ars Electronica website (https://ars.electronica.art/prix/en/digitalhumanity/).
2025.10.13 View 3522 -
City AI Research Institute Selected for Ministry of Science and ICT's Brain Pool (BP) Institutional Recruitment Program
<Professor Yoonjin Yoon from the Department of Civil and Environmental Engineering at KAIST>
KAIST's City AI Research Institute (Director: Professor Yoonjin Yoon) has been selected for the Ministry of Science and ICT's Brain Pool (BP) Institutional Recruitment Program. This achievement is the culmination of a joint proposal spearheaded by Institute Director Professor Yoonjin Yoon, along with Professor Soyoung In of the Department of Civil and Environmental Engineering and Professor Sujin Han of the School of Electrical Engineering. It is the result of high praise for the institute's research capabilities in the field of Urban AI and its potential for international collaboration.
This BP project, with a total budget of 2.1 billion KRW, will be carried out over 28 months. It plans to actively pursue AI research focused on solving urban problems by inviting renowned overseas scholars to focus on three core areas: Geospatial AI, Climate AI, and Physical AI. Through this, the institute aims to develop core AI technologies based on a collaboration system involving industry, academia, research institutions, and government. This will lead the way in sustainable urban growth and the transition to an 'Cognitive City,' continuing research to proactively diagnose and respond to various issues that citizens can experience firsthand.
This project is particularly significant as it is a female-centered institutional Brain Pool project. KAIST plans to systematically support the growth of early-career female researchers and actively expand the participation of next-generation female scientists and engineers in international research networks. This is expected to significantly contribute to the development of female research personnel and the strengthening of research leadership, areas that are relatively lacking in domestic science and engineering fields.
Furthermore, through long-term joint research with researchers from world-leading universities such as MIT, NYU, UIUC, UBC, USF, and the University of Toronto, the City AI Research Institute is set to become a leading Urban AI research hub in Korea and Asia.
Moving forward, the institute will continue to dedicate itself to core research for responding to the complex challenges of future cities and advancing innovative technology through artificial intelligence, based on global cooperation.
2025.10.10 View 1843 -
Physics Informed AI Excels at Large Scale Discovery of New Materials!
<(From left) Ph.D candidates Songho Lee, Donggeun Park, and Hyeonbin Moon, and Professor Seunghwa Ryu from the Department of Mechanical Engineering; (top) Professor Jae Hyuk Lim from Kyung Hee University and Dr. Wabi Demeke from KAIST>
One of the key steps in developing new materials is “property identification,” which has long relied on massive amounts of experimental data and expensive equipment, limiting research efficiency. A KAIST research team has introduced a new technique that combines “physical laws,” which govern deformation and interaction of materials and energy, with artificial intelligence. This approach allows for rapid exploration of new materials even under data-scarce conditions and provides a foundation for accelerating design and verification across multiple engineering fields, including materials, mechanics, energy, and electronics.
KAIST (President Kwang Hyung Lee) announced on the 2nd of October that Professor Seunghwa Ryu’s research group in the Department of Mechanical Engineering, in collaboration with Professor Jae Hyuk Lim’s group at Kyung Hee University (President Jinsang Kim) and Dr. Byungki Ryu at the Korea Electrotechnology Research Institute (President Namkyun Kim), proposed a new method that can accurately determine material properties with only limited data. The method uses Physics-Informed Machine Learning (PIML), which directly incorporates physical laws into the AI learning process.
<Schematic Diagram of a Physics-Based Machine Learning Methodology for Understanding Material Properties>
In the first study, the researchers focused on hyperelastic materials, such as rubber. They presented a Physics-Informed Neural Network (PINN) method that can identify both the deformation behavior and the properties of materials using only a small amount of data obtained from a single experiment. Whereas previous approaches required large, complex datasets, this research demonstrated that material characteristics can be reliably reproduced even when data is scarce, limited, or noisy.
In the second study, the team turned to thermoelectric materials—new materials that convert heat into electricity and electricity into heat. They proposed a PINN-based inverse inference technique that can estimate key indicators, such as thermal conductivity (how well heat is transferred) and the Seebeck coefficient (how efficiently electricity is generated), from just a few measurements.
Going further, the researchers introduced a Physics-Informed Neural Operator (PINO), an AI model that understands the physical laws of nature, and showed that it can generalize to previously unseen materials without requiring retraining.
In fact, after training the system on 20 materials, they tested it on 60 entirely new materials, and in all cases it predicted their properties with high accuracy. This breakthrough points to a future where large-scale, high-speed screening of countless candidate materials becomes possible.
This achievement goes beyond simply reducing the need for experiments. By intricately combining physical laws with AI, the researchers provided the first example of improving experimental efficiency while preserving reliability.
Professor Seunghwa Ryu, who led both studies, stated, “This is the first case of applying AI that understands physical laws to real material research. It enables reliable identification of material properties even when data availability is limited, and it is expected to expand into various engineering fields.”
The first paper, co-first-authored by KAIST Mechanical Engineering PhD candidates Hyeonbin Moon and Donggeun Park, was published on August 13 in Computer Methods in Applied Mechanics and Engineering.
※ Paper title: “Physics-informed neural network-based discovery of hyperelastic constitutive models from extremely scarce data”
※ DOI: https://doi.org/10.1016/j.cma.2025.118258
The second paper, co-first-authored by KAIST Mechanical Engineering PhD candidates Hyeonbin Moon and Songho Lee, and Dr. Wabi Demeke, was published on August 22 in npj Computational Materials.
※ Paper title: “Physics-informed neural operators for generalizable and label-free inference of temperature-dependent thermoelectric properties” ※ DOI: https://doi.org/10.1038/s41524-025-01769-1
Meanwhile, the first study was supported by the Korea Research Foundation and the Ministry of Science and ICT’s INNOCore Program, as well as by a research project from the Ministry of Food and Drug Safety. The second study was carried out with support from the Korea Research Foundation and the Ministry of Science and ICT’s INNOCore Program.
2025.10.10 View 2447 -
Next Generation Robots Roaming Shipyards and City Centers
< Diden Robotics Research Team Co., Ltd (Leftmost person in the front row is CEO Joon-Ha Kim)>
KAIST announced on the September 30th that domestic robot startups, founded on KAIST research achievements, are driving new innovation at shipyards and urban worksites.
An industrial walking robot that freely climbs walls and ceilings and a humanoid walking robot that walks through downtown Gangnam are attracting attention as they enter the stage of commercialization. The stars are DIDEN Robotics Co., Ltd. and Eurobotics Co., Ltd.
Diden Robotics is providing a new breakthrough in the industrial automation market, including the shipbuilding industry, by commercializing its innovative 'Seungwol (Ascend and Cross) Robot' technology, which allows it to move freely and work on steel walls and ceilings. Eurobotics is commercializing world-class humanoid walking technology, and this achievement is scheduled to be officially presented at the international humanoid robot conference, 'Humanoids 2025,' to be held on October 1st.
< Diden Robot's Outer Plate (Longi) and Welding Test >
Diden Robotics is a robotics startup jointly founded in March 2024 by four alumni from the KAIST Mechanical Engineering Hu-bo Lab DRCD research team (Professor Hae-Won Park). Its flagship product, 'DIDEN 30,' is a quadrupedal robot designed for use in high-risk work environments that are difficult for humans to access, combining autonomous driving technology, a foot-shaped leg structure, and magnetic feet.
The 'DIDEN 30' successfully completed the 'Longitudinal (longi) Overcoming Test,' in which it stepped over steel stiffeners (longitudinals) densely installed as part of the structure at a ship construction site, proving its potential for field deployment. Currently, the company is conducting research to enhance its functionality so it can stably pass through access holes, the narrow entryways inside ships. It is also pushing for performance improvements so it can be deployed for real tasks such as welding, inspection, and painting starting in the second half of 2026.
A next-generation bipedal walking robot, 'DIDEN Walker,' is also under development. Targeting the completion of a prototype in the fourth quarter of 2025, it is being designed for stable walking in cramped and complex industrial environments. Plans are also underway to equip it with an upper-body manipulator for automated welding in the shipbuilding industry.
Diden Robotics is accelerating the advancement of its proprietary 'Physical AI' technology. The core is the self-developed AI learning platform, 'DIDEN World,' which applies an offline reinforcement learning method where the AI generates optimal motion data in a virtual simulation beforehand and learns without trial and error, increasing learning efficiency and stability.
< Diden Robot (DIDEN 30) >
Furthermore, to actually implement the AI technology, the company is internalizing its hardware and advancing its 3D recognition technology, which serves as the robot's 'eyes.' It is aiming for a completely autonomous walking system that requires no worker intervention by 2026, using technology such as 3D mapping based on four cameras.
In addition to this technological development, Diden Robotics successfully performed the longitudinal overcoming, Seungwol test, and welding work on blocks under construction through a joint development with Samsung Heavy Industries in September. This is a significant achievement, meaning Diden Robotics' technology has been validated in actual industrial settings, moving beyond the laboratory level.
Meanwhile, Diden Robotics is collaborating with major domestic shipyards, including Samsung Heavy Industries, HD Hyundai Samho, Hanwha Ocean, and HD Korea Shipbuilding & Offshore Engineering, to develop site-customized robots.
Joon-Ha Kim, CEO of Diden Robotics, stated, "The successful tests at the Samsung Heavy Industries site proved the practicality and stability of our technology. We will establish ourselves as a leading company in solving labor shortages and driving automation in the shipbuilding industry."
< (Eurobotics Research Team Co., Ltd.)(Leftmost person in the top row is CEO Byung-ho Yoo) >
Eurobotics is an autonomous walking startup jointly founded by three alumni from Professor Hyun Myung's research team at KAIST. It is promoting the commercialization of autonomous walking technology for indoor and outdoor industrial sites, including rough terrain. In a recently released video, a humanoid equipped with control technology developed by Eurobotics attracted attention by walking naturally through the crowd in downtown Gangnam.
The core technology is the 'Blind Walking Controller.' It determines locomotion based only on internal information without external sensors like cameras or LiDAR, enabling stable walking regardless of day, night, or weather. The robot performs locomotion by 'imagining' the terrain without precise terrain modeling, demonstrating robust performance with the same controller across various environments such as sidewalks, downhill slopes, and stairs.
This technology originated from the quadrupedal walking competition at the 2023 International Conference on Robotics and Automation (ICRA), where Professor Myung's lab participated, and proved its world-class capability by winning, beating MIT by a large margin. At the time, Byungg-ho Yoo, CEO of Eurobotics, led the team, and Co-CTOs Min-ho Oh and Dong-kyu Lee directly participated in developing the core autonomous walking technology. Based on this, they continued further development tailored to the humanoid environment and have entered the commercialization stage.
< Eurobotics' Humanoid Walking >
Byung-ho Yoo, CEO of Eurobotics, emphasized, "This video is the first step toward complete humanoid autonomous walking. We will develop KAIST's research achievements into technologies that can be immediately utilized in industrial settings."
Hyeonmin Bae, Head of the KAIST Startup Center, said, "We will provide close support from the initial stages to help the on-campus robotics industry grow actively and assist them in settling down stably."
Kwang Hyung Lee, President of KAIST, stated, "This achievement is a representative case showing that KAIST's fundamental technologies are rapidly spreading to industrial fields through startups. KAIST will continue to actively support innovative entrepreneurship based on challenging research and help lead the global robotics industry."
※ https://2025humanoids.org https://www.seoulairobot.com/
2025.09.30 View 5300 -
3D Printing Becomes Stronger and More Economical with Light and AI
<(Front) Ph.D. candidate Jisoo Nam, (Back row, from left) Ph.D. candidate Boxin Chen, Professor Miso Kim>
Photocurable 3D printing, widely used for everything from dental treatments to complex prototype manufacturing, is fast and precise but has the limitation of being fragile and easily broken by impact. A KAIST research team has developed a new technology to overcome this weakness, paving the way for the more robust and economical production of everything from medical implants to precision machine parts.
KAIST (President Kwang Hyung Lee) announced on the 29th that Professor Miso Kim's research team in the Department of Mechanical Engineering has developed a new technology that fundamentally resolves the durability limitations of photocurable 3D printing.
Digital Light Processing (DLP)-based 3D printing is a technique that uses light to solidify liquid resin (polymer) to rapidly manufacture precise structures, used in various fields such as dentistry and precision machinery. While traditional injection molding offers excellent durability, it requires significant time and cost for mold fabrication. In contrast, photocurable 3D printing allows for flexible shape realization but has a durability drawback.
Professor Kim's team solved this problem by combining two key elements:
A new photocurable resin material that absorbs shock and vibration while allowing for a wide range of properties from rubber to plastic.
A machine learning-based design technology that automatically assigns optimal strength to each part of the structure.
<Figure 1. Schematic of a new manufacturing technology for high-durability photocurable 3D printing using light-controlled gradient structures. This approach integrates the development of stiffness-controllable viscoelastic polyurethane acrylate (PUA) materials, machine learning-based property gradient optimization, and grayscale DLP 3D printing. The technology enhances damping performance and alleviates stress concentration, providing an integrated solution for high reliability, durability, and customized manufacturing. It demonstrates potential applications in structural components subjected to repetitive loads such as joints, automotive interior parts, and precision machinery components>
The research team developed a Polyurethane Acrylate (PUA) material incorporating dynamic bonds, which significantly increases shock and vibration absorption capability compared to existing materials. Furthermore, they successfully applied 'grayscale DLP' technology, which controls the light intensity to achieve different strengths from a single resin composition, thereby assigning customized strength to specific areas within the structure. This concept is inspired by the harmonious and different roles played by bone and cartilage in the human body.
A machine learning algorithm automatically proposes the optimal strength distribution by analyzing the structure and load conditions. This organically connects material development and structural design, enabling customized strength distribution.
The economic efficiency is also noteworthy. Previously, expensive 'multi-material printing' technology was required to achieve diverse material properties, but this new technology yields the same effect with a single material and a single process, significantly reducing production costs. It eliminates the need for complex equipment or material management, and the AI-based structural optimization shortens research and development time and product design costs.
Professor Miso Kim explained, "This technology simultaneously expands the degrees of freedom in material properties and structural design. Patient-specific implants will become more durable and comfortable, and precision machine parts can be manufactured more robustly." She added, "The fact that it secures economic viability by realizing various strengths with a single material and single process is highly significant," and "We anticipate its utilization across various industrial fields such as biomedical, aerospace, and robotics."
The research was spearheaded by Professor Miso Kim's team at the KAIST Department of Mechanical Engineering, with Ph.D. candidate Jisoo Nam as the first author. Boxin Chen, a student from Sungkyunkwan University, also contributed to the collaborative research. The findings were published online on July 16 in the world-renowned journal in materials science, Advanced Materials (IF 26.8). Recognizing the research's excellence, it was also selected for the journal's Frontispiece.
Paper Title: Machine Learning-Driven Grayscale Digital Light Processing for Mechanically Robust 3D-Printed Gradient Materials
DOI: 10.1002/adma.202504075
The achievements of this research have brought Professor Miso Kim significant international attention, as she simultaneously received the 'Wiley Rising Star Award' and the 'Wiley Women in Materials Science Award' in July 2025, hosted by the international academic publisher Wiley.
The Wiley Rising Star Award is given to emerging researchers with the potential for academic leadership, and the Wiley Women in Materials Science Award is a prestigious honor established to celebrate outstanding female scientists in the field of materials science.
<Figure 2. Frontispiece image (scheduled for Issue 42). Multi-property structure fabricated using a photocurable 3D printer. By varying the projector light intensity by location, stronger light creates rigid regions while weaker light forms flexible ones. AI designs an optimized pattern for the structural shape to prevent fracture and reinforce the overall strength.>
This research was supported by the National Research Foundation of Korea (NRF) grants funded by the Korean government (MSIT) (Nos. NRF-2021R1A2C2095767, RS-2023-00254689, and RS-2024-00433654).
2025.09.30 View 3080 -
KAIST Develops Semiconductor Neuron that Remembers and Responds Like the Brain
<(From left, clockwise) Professor Kyung Min Kim, Min-Gu Lee, Dae-Hee Kim, Dr. Han-Chan Song, Tae-Uk Ko, Moon-Gu Choi, and Eun-Young Kim>
The human brain does more than simply regulate synapses that exchange signals; individual neurons also process information through “intrinsic plasticity,” the adaptive ability to become more sensitive or less sensitive depending on context. Existing artificial intelligence semiconductors, however, have struggled to mimic this flexibility of the brain. A KAIST research team has now developed next-generation, ultra-low-power semiconductor technology that implements this ability as well, drawing significant attention.
KAIST (President Kwang Hyung Lee) announced on September 28 that a research team led by Professor Kyung Min Kim of the Department of Materials Science and Engineering developed a “Frequency Switching Neuristor” that mimics “intrinsic plasticity,” a property that allows neurons to remember past activity and autonomously adjust their response characteristics.
“Intrinsic plasticity” refers to the brain’s adaptive ability- for example, becoming less startled when hearing the same sound repeatedly, or responding more quickly to a specific stimulus after repeated training. The “Frequency Switching Neuristor” is an artificial neuron device that autonomously adjusts the frequency of its signals, much like how the brain becomes less startled by repeated stimuli or, conversely, increasingly sensitive through training.
The research team combined a “volatile Mott memristor,” which reacts momentarily before returning to its original state, with a “non-volatile memristor,” which remembers input signals for long periods of time. This enabled the implementation of a device that can freely control how often a neuron fires (its spiking frequency).
<Figure 1. Conceptual comparison between a neuron and a frequency-tunable neuristor. The intrinsic plasticity of brain neurons regulates excitability through ion channels. Similarly, the frequency-tunable neuristor uses a volatile Mott device to generate current spikes, while a non-volatile VCM device adjusts resistance states to realize comparable frequency modulation characteristics>
In this device, neuronal spike signals and memristor resistance changes influence each other, automatically adjusting responses. Put simply, it reproduces within a single semiconductor device how the brain becomes less startled by repeated sounds or more sensitive to repeated stimuli.
To verify the effectiveness of this technology, the researchers conducted simulations with a “sparse neural network.” They found that, through the neuron’s built-in memory function, the system achieved the same performance with 27.7% less energy consumption compared to conventional neural networks.
They also demonstrated excellent resilience: even if some neurons were damaged, intrinsic plasticity allowed the network to reorganize itself and restore performance. In other words, artificial intelligence using this technology consumes less electricity while maintaining performance, and it can compensate for partial circuit failures to resume normal operation.
Professor Kyung Min Kim, who led the research, stated, “This study implemented intrinsic plasticity, a core function of the brain, in a single semiconductor device, thereby advancing the energy efficiency and stability of AI hardware to a new level. This technology, which enables devices to remember their own state and adapt or recover even from damage, can serve as a key component in systems requiring long-term stability, such as edge computing and autonomous driving.”
This research was carried out with Dr. Woojoon Park (now at Forschungszentrum Jülich, Germany) and Dr. Hanchan Song (now at ETRI) as co-first authors, and the results were published online on August 18 in Advanced Materials (IF 26.8), a leading international journal in materials science.
※ Paper title: “Frequency Switching Neuristor for Realizing Intrinsic Plasticity and Enabling Robust Neuromorphic Computing,” DOI: 10.1002/adma.202502255
This research was supported by the National Research Foundation of Korea and Samsung Electronics.
2025.09.30 View 2713 -
Thinking outside the box to Fabricate Customized 3D Neural Chips
<(From Left) Professor Yoonkey Nam, Dr. Dongjo Yoon from the Department of Bio and Brain Engineering>
Cultured neural tissues have been widely used as a simplified experimental model for brain research. However, existing devices for growing and recording neural tissues, which are manufactured using semiconductor processes, have limitations in terms of shape modification and the implementation of three-dimensional (3D) structures.
By "thinking outside the box," a KAIST research team has successfully created a customized 3D neural chip. They first used a 3D printer to fabricate a hollow channel structure, then used capillary action to automatically fill the channels with conductive ink, creating the electrodes and wiring. This achievement is expected to significantly increase the design freedom and versatility of brain science and brain engineering research platforms.
On the 25th, KAIST announced that a research team led by Professor Yoonkey Nam from the Department of Bio and Brain Engineering has successfully developed a platform technology that overcomes the limitations of traditional semiconductor-based manufacturing. This technology allows for the precise fabrication of "3D microelectrode array" (neural interfaces with multiple microelectrodes arranged in a 3D space to measure and stimulate the electrophysiological signal of neurons) in various customized forms for in vitro culture chips.
Existing 3D microelectrode array fabrication, based on semiconductor processes, has limited 3D design freedom and is expensive. While 3D printing-based fabrication techniques have recently been proposed to overcome these issues, they still have limitations in terms of 3D design freedom for various in vitro neural network structures because they follow the traditional sequence of "conductive material patterning → insulator coating → electrode opening."
The KAIST research team leveraged the excellent 3D design freedom provided by 3D printing technology and its ability to use printed materials as insulators. By reversing the traditional process, they established an innovative method that allows for more flexible design and functional measurement of 3D neuronal network models for in vitro culture.
<Schematic Diagram of an Integrated Cell Culture Substrate-Microelectrode Array Platform for In Vitro Cultured 3D Neural Network Models>
First, they used a 3D printer to print a hollow 3D insulator with micro-tunnels. This structure was designed to serve as a stable scaffold for conductive materials in 3D space while also supporting the creation of various 3D neuronal networks. They then demonstrated that by using capillary action to fill these internal micro-tunnels with conductive ink, they could create a 3D scaffold-microelectrode array with more freely arranged microelectrodes within a complex 3D culture support structure.
The new platform can be used to create various chip shapes, such as probe-type, cube-type, and modular-type, and supports the fabrication of electrodes using different materials like graphite, conductive polymers, and silver nanoparticles. This allows for the simultaneous measurement of multichannel neural signals from both inside and outside the 3D neuronal network, enabling precise analysis of the dynamic interactions and connectivity between neurons.
Professor Nam stated, "This research, which combines 3D printing and capillary action, is an achievement that significantly expands the freedom of neural chip fabrication." He added that it will contribute to the advancement of fundamental brain science research using neural tissue, as well as applied fields like cell-based biosensors and biocomputing.
Dr. Dongjo Yoon from KAIST's Department of Bio and Brain Engineering participated as the first author of the study. The research findings were published online in the international academic journal Advanced Functional Materials (June 25th issue).
※Paper Title: Highly Customizable Scaffold-Type 3D Microelectrode Array Platform for Design and Analysis of the 3D Neuronal Network In Vitro
This research was supported by the Consolidator Grants Program and the Global Basic Research Laboratory Program of the National Research Foundation of Korea.
2025.09.26 View 2276 -
KAIST team links early life epigenetic memory to adult brain inflammation
<(From left) Professor Won-Suk Chung, Ph.D. Ph.D candidate Hyeonji Park Dr. Seongwan Park, Professor Inkyung Jung>
Why do some people remain healthy through childhood yet become more vulnerable to brain disorders such as dementia later in life? A KAIST (President Kwang Hyung Lee) -led team has uncovered a key part of the answer: a developmental ‘switch’ in astrocytes—the brain’s most abundant support cells that shapes how strongly the brain’s immune system reacts in adulthood. The study identifies a gene, NR3C1 (encoding the glucocorticoid receptor), as a master regulator of this switch and shows how early-life epigenetic ‘memory’ can predispose the adult brain to excessive inflammation.
The work was carried out by a joint team led by Professor Inkyung Jung (Department of Biological Sciences, KAIST) and Associate Director Won-Suk Chung (Center for Vascular Research, Institute for Basic Science; Professor, KAIST Biological Sciences). Using mouse models, the researchers mapped gene-regulatory programs across multiple stages of astrocyte development and found that NR3C1 acts during a brief early-postnatal window to enforce long-term immune restraint.
<The schematic illustrates how the NR3C1 gene (glucocorticoid receptor) suppresses the immune response of astrocytes. In normal (control) astrocytes, NR3C1 binds to specific regulatory regions of DNA (nGRE) to inhibit the expression of immune-related genes, thereby maintaining brain homeostasis even under immune stimulation. In contrast, in NR3C1-deficient astrocytes (KO), this suppression is lost, leading to excessive activation of inflammation-related genes such as Gfap, Il6st, Stat2, and Cxcl10. As a result, in an autoimmune encephalomyelitis (EAE) model, pronounced neuroinflammation and clinical symptoms (paralysis and severe debilitation) are observed>
To build this map, the team combined state-of-the-art 3D epigenome profiling with RNA sequencing and chromatin accessibility analyses, capturing how DNA folds and which regulatory elements contact target genes. They identified 55 stage-specific transcription factors that guide astrocyte maturation; among them, NR3C1 emerged as the critical ‘switch’ in early life. Notably, deleting NR3C1 in astrocytes did not disrupt normal development. However, when the adult mice were challenged with an autoimmune model of multiple sclerosis, animals lacking astrocytic NR3C1 mounted exaggerated inflammatory responses and developed more severe disease.
Mechanistically, the study shows that early loss of NR3C1 epigenetically primes immune genes - keeping their regulatory elements open and ready - so that later in life these genes respond too strongly to inflammatory cues. In effect, NR3C1 serves as an early ‘brake’ that prevents over-activation of astrocyte immune programs in adulthood.
“This is the first demonstration that astrocyte immune functions are governed by epigenetic memory,” said Professor Won-Suk Chung. “Our findings offer new clues to the origins of degenerative brain disorders, including Alzheimer’s disease.”
“We reveal a temporal regulatory window in astrocyte development that can set the stage for disease vulnerability in adulthood,” added Professor Inkyung Jung. “Understanding the 3D genome logic behind these programs could open paths to therapies for immune-related brain disorders such as multiple sclerosis.”
<The figure shows the three-dimensional genome structure of astrocytes at specific gene loci, illustrating how NR3C1 regulates their expression. In normal cells, NR3C1 binds to DNA and maintains the chromatin in a closed state, thereby preventing unnecessary activation between distal regulatory elements (enhancers) and gene promoters. In contrast, when NR3C1 is absent, the chromatin becomes open, creating a state in which enhancers and genes can be more easily activated. As a result, genes such as Mxi1 are overexpressed, triggering inflammatory responses. This clearly demonstrates that NR3C1 plays an essential role in maintaining immune homeostasis by stabilizing three-dimensional gene regulatory mechanisms.>
The results of this study were published online on September 22 in the international journal Nature Communications (IF 15.7), with Dr. Seongwan Park and PhD student Hyeonji Park of KAIST’s Department of Biological Sciences as co-first authors.
※ Paper title: “NR3C1-mediated epigenetic regulation suppresses astrocytic immune responses in mice,” DOI: https://www.nature.com/articles/s41467-025-64088-5
In addition, on September 17, the journal published a commentary article introducing this research: https://www.nature.com/articles/s41467-025-64102-w
This research was supported by the Suh Kyungbae Science Foundation, the Ministry of Health and Welfare, the Ministry of Science and ICT, and IBS.
Glossary - Epigenetic priming: preparing genes for rapid future activation by altering chromatin without changing DNA sequence
2025.09.25 View 3143