Deepnoid Improves Medical Optimization AI Solution Using DeepSeek
Medical AI company Deepnoid announced on the 5th that it will apply DeepSeek's inference model (R1) to its proprietary generative LLMOps platform 'DEEP:GEN'.
DEEP:GEN is a medical-optimized AI service platform that selectively incorporates the advantages of various open sources such as LLaMA and DeepSeek. By applying DeepSeek R1, which features reinforcement learning and knowledge distillation, the accuracy of inference and generation has been improved.
With R1 applied, DEEP:GEN learns to enable optimal medical decision-making even with limited data and continuous environmental changes. When DEEP:GEN is used at Hospital A, the AI autonomously learns according to Hospital A’s medical environment and progressively optimizes itself.
DEEP:GEN is lightweighted through knowledge distillation. This allows the knowledge of high-performance AI models to be transferred to smaller AI models, optimizing to maintain high accuracy even on low-spec servers. It enables consistent performance across different IT infrastructure environments in hospitals, allowing DEEP:GEN to be used quickly and efficiently in cloud environments as well as with various medical devices and PACS systems.
A thorough security system is established. Since it operates in a cloud environment physically separated from DeepSeek, user input data is not leaked to any specific country. The data is not used for model training by the R1 developer.
Senior Researcher Byeongcheol Yoo of Deepnoid said, "We plan to build vertical integration of medical AI solutions with DEEP:GEN and capture the domestic and international remote reading service markets," adding, "We will expand AI-based medical services such as clinical support and disease prediction."
Director Sukyung Kim of Deepnoid emphasized, "AI will become a key tool that assists doctors’ clinical practice to realize precise and personalized medical services," and added, "It is expected to contribute to strengthening medical staff’s decision-making, improving work efficiency, and supporting patient-tailored treatment."
Hot Picks Today
"Rather Than Endure a 1.5 Million KRW Stipend, I'd Rather Earn 500 Million in the U.S." Top Talent from SNU and KAIST Are Leaving [Scientists Are Disappearing] ①
- "You Might Regret Not Buying Now"... Overseas Retail Investors Stirred by News of Record-Breaking Monster Stocks' IPOs
- "Not Jealous of Winning the Lottery"... Entire Village Stunned as 200 Million Won Jackpot of Wild Ginseng Cluster Discovered at Jirisan
- Shinsegae Vice President Visits May 18 Bereaved Families, Apology for 'Tank Day' Controversy Rejected: "Will Apologize Again After Full Investigation"
- "How Did an Employee Who Loved Samsung End Up Like This?"... Past Video of Samsung Electronics Union Chairman Resurfaces
Deepnoid plans to expand the application of DEEP:GEN equipped with DeepSeek to the fields of security AI and machine vision.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.