[AI Revolution](148) Creating "My Own AI Health Assistant" by 'Sagak'

No Worries About Health Data Leaks
Reads National Health Insurance and Smartwatch Data
Personalized Analysis of Sleep, Heart Rate, and Exercise

Baek Geon-gyu, CEO of Sagak, is introducing the self-developed AI model 'Candy AI' in an interview with Asia Economy. Photo by Park Yujin

When users input the command "Analyze my health data" and enter their authentication information, health information such as height, weight, and blood pressure registered with the National Health Insurance Service appears on the screen. Based on this data, artificial intelligence (AI) recommends nutritional supplements. Additionally, upon the request "Analyze and summarize my smartwatch data," personalized analyses of sleep, heart rate, and exercise volume are provided based on data from connected wearable devices.

'Sagak,' a deep-tech startup founded by CEO Bae Geon-gyu, who worked at Samsung Electronics for 12 years, developed the ultra-personalized AI solution 'CANDiY AI' that utilizes personal health information. Having just passed its third year since establishment, Sagak's core technology is 'on-device AI,' which provides customized health services while ensuring that users' sensitive health data does not leave their devices.

The biggest differentiator of the solution developed by Sagak is its method of analyzing personal health data from sources such as the National Health Insurance Service and smartwatches within the user's device. This technology, known as a 'small language model (sLM)' or 'vertical AI,' is smaller than general AI models but specialized in specific fields, enabling it to run on personal devices.

Explanation of the solution 'CandyAI' developed by 'Sagak'. Photo by Sagak

On the 11th, during a meeting with Asia Economy at Seoul Startup Hub Gongdeok, CEO Bae said, "ChatGPT is a massive model with over one trillion parameters (where AI stores learned data), but we use a smaller model with 8 billion parameters. This allows it to run on smartphones." He added, "This small model is specialized in healthcare, analyzing users' health data to provide customized health advice."

Sagak has already conducted pilot services in South Korea with Asan Medical Center and Cha Hospital, offering drug and supplement recommendations along with personalized advice, achieving participant satisfaction rates of 96% and 93.7%. If non-face-to-face medical consultations become possible domestically, they plan to offer services that connect patients with doctors remotely.

At this year's Mobile World Congress (MWC) held in Barcelona, Spain, Sagak participated in the startup competition '4YFN' hosted by the Global System for Mobile Communications Association (GSMA) and ranked within the top 20. Subsequently, Sagak is set to be the first Korean startup listed in GSMA's 'AI Use Case Library.' This library is a platform showcasing real-world AI use cases to telecom companies worldwide and had previously only featured global giants such as Huawei, Ericsson, IBM, China Mobile, and Nokia.

CEO Bae stated, "In the future, all applications will disappear, leaving only AI assistants. For example, a world will come where you say, 'Agent, transfer money from Hana Bank to Kookmin Bank,' and it is processed immediately. We aim to be the best in the healthcare sector among them."

Sagak expects full-scale sales starting from the third quarter of this year, setting a minimum sales target of 3 billion KRW. They are also preparing for Series A investment rounds around the same time. He said, "Since we have confirmed high demand in overseas markets where non-face-to-face medical consultations are possible, we aim to list on NASDAQ within five years."