On September 14, the Korea Aerospace Research Institute (KARI) announced the launch of an AI Transformation (AX) initiative for its research and administrative operations, with plans to develop an AI platform named "KARI BRAIN." This platform aims to integrate the institute's extensive aerospace research and development knowledge and data with AI systems. According to current plans, the platform is targeted to be operational by the end of 2027.
On the same day, KARI held an AX Leaders Launch Meeting, introducing 13 selected AX leaders to the future direction and specific responsibilities of the initiative. These individuals, who are part of the AX Promotion Office established in July this year, will focus on identifying AI application scenarios that meet the actual needs of researchers and business units while continuing to perform their existing duties.
Korea Aerospace Research Institute Plans to Launch AI Platform by 2027
In fact, KARI's AI development has been ongoing for some time. The institute established a dedicated AI infrastructure task force in May 2024 and deployed a server equipped with four NVIDIA H200 GPUs in January this year. In April, it added another server with eight NVIDIA B200 GPUs to provide computational support for internal AI applications.
Currently, the AI models used by KARI have a total of 320 billion parameters, with approximately 18 billion activated parameters involved in each actual computation. The model supports multimodal inputs such as images, screenshots, and documents, with application focuses including code writing, intelligent agent tasks, document processing, and visual understanding. Meanwhile, the institute's generative AI service now has 836 users, and AI is gradually transitioning from an experimental tool to an integral part of actual workflows.
From a practical application perspective, KARI plans to first expand the use of AI in research and development, including technical data analysis, design and engineering analysis support, test data processing, and software development. The goal is to leverage AI to reduce the burden of repetitive tasks and data processing. In the administrative field, the institute will explore using AI for regulatory retrieval, document writing, and automating repetitive tasks, thereby allowing researchers to spend less time on non-core activities and focus more on scientific research.
In terms of implementation pace, KARI has designated the period from August 2026 to July 2027 as the first phase of AX implementation. During this time, the institute will evaluate the effectiveness of AI tools and usage methods through practical applications and further promote them across the entire organization. This means that its AI transformation is not just about deploying models but aims to verify whether AI can truly improve research and administrative efficiency through real-world business scenarios.
Additionally, data security and the reliability of AI outputs are listed as key priorities. KARI is currently providing tools for building personal RAG (Retrieval-Augmented Generation) systems and encouraging researchers to organize and create data suitable for AI use. As relevant data is gradually formed, the institute plans to set corresponding security levels for different AI data and use incentive mechanisms to promote continuous data collection, ultimately forming public data assets that can be shared and reused within the organization.