Acryl Co., Ltd. (CEO Park Wae-jin), a company specializing in AI platforms, announced on the 1st that it has signed a memorandum of understanding (MOU) with Yonsei University Digital Health Institute (Director Kim Hyun-chang) for joint research cooperation on a globally top-tier 'medical-specialized foundation model'.
This agreement is significant as it aims to promote the practical industrialization of medical AI by combining Acryl's AI technology with the high-quality clinical data (CDW) possessed by Yonsei University Severance Hospital. Particularly, data consistency and terminology standardization based on a single institution, along with long-term follow-up medical records, boast the highest quality in the country. It is expected to contribute to enhancing the strategic value of national medical data assets by being utilized in disease prognosis prediction, survival rate analysis, and early detection of high-risk groups.
Through this collaboration, Acryl plans to officially develop healthcare-specialized services by integrating agentic AI technology based on a large-scale language model operating system (LLMOps). The company intends to expand the scope of AI utilization beyond diagnosis to personalized health promotion and disease prevention, and to spread federated learning across the industry, leading innovation in the medical AI ecosystem.
Park Wae-jin, CEO of Acryl, stated, "To become one of the top three AI powerhouses, a medical-specialized foundation model based on Korea's strength in medical data is necessary, as a general-purpose foundation model alone is insufficient." He added, "Korea possesses a unique competitive edge with the world's highest level of medical data digitalization and a single insurance system, combining data consistency and scalability." Park further mentioned, "Acryl will build a practical medical AI ecosystem centered on hospital data on this foundation to secure top-tier global competitiveness in medical AI."
Meanwhile, Acryl's AAAI research institute recently proved its technological prowess in the medical AI field by advancing to the finals as one of the top 20 teams out of 630 research teams in the 'AI Champion Contest' hosted by the Ministry of Science and ICT.
The 'multi-agent workflow and service based on generative AI specialized in medical record automation' submitted in the contest is a technology that automates and intelligentizes the medical field's clinical processes. It is regarded as a core foundation for the development of the medical-specialized foundation model that Acryl is pursuing. This model is a lightweight foundation model with 4 billion parameters, trained by combining Acryl's medical dataset with publicly available medical datasets, and applies a self-developed 'multi-layer hallucination suppression' structure. Thanks to its lightweight structure, it can operate independently within hospitals without large-scale GPU bases, and through multi-layer hallucination suppression technology, it achieves response quality close to that of ultra-large models (over 100 billion parameters).
Through this, Acryl's model, which has secured practicality, reliability, and cost efficiency, is evaluated as a next-generation AI solution that can be immediately utilized in medical fields. Acryl plans to accelerate the enhancement of its medical-specialized foundation model, combining medical data and generative AI, to secure global-level competitiveness in medical AI.
ⓒ dongA.com. All rights reserved. Reproduction, redistribution, or use for AI training prohibited.
Popular News