Miso Information Technology: “Garbage Data Is Useless for AI — 20 Years of Data Expertise to Drive Multi-Hundred-Billion-Won Growth”
It is now widely accepted that data sits at the core of the AI transition era. At the same time, data is a double-edged sword. As seen during the earlier big data boom, large volumes of poorly managed data can become nothing more than a “mountain of trash” for AI systems. When trained on flawed data, AI may produce results that appear plausible but are fundamentally incorrect—often at critical moments. Despite the nationwide push for AI transformation (AX), however, few organizations today possess data that is truly ready and reliable for AI use.
Against this backdrop, Miso Information Technology, a company that has quietly focused on data for more than 20 years, is being newly reassessed by the industry. The company has drawn particular attention for delivering tangible results in fields that demand especially high levels of AI reliability, such as healthcare and smart factories. In July, Miso Information Technology appointed Nam Sang-do—a technology-and-business hybrid leader who has served as a founding member, CTO, and COO—as its new CEO. Since then, the company has accelerated growth around an “AI data–centric” strategy. We met with CEO Nam to discuss the competitiveness of data-driven companies in the AI era and the company’s vision for a 2026 IPO.
“You Can’t Put Everything into One Container”: The Power of Fabric Architecture
“We have consistently focused on data—nothing else,” Nam said. “From the early days through the data lake boom, when big data became a major trend, we concentrated on transforming diverse datasets into usable assets. Now, as we enter the full AI era, awareness of data’s importance is spreading rapidly across industries. For us, this shift represents a major growth opportunity, because we already provide one-stop support from long-accumulated data to AI operations.”
Rather than highlighting short-term results driven by the current AI hype, Nam emphasized the long-term experience and expertise gained by “digging a single well” for decades. A key advantage, he noted, is the company’s early development of proprietary platforms for MLOps (machine learning operations) and LLMOps (large language model operations), which are now essential components of modern AI systems.
In the AI era, data reliability has become more critical than ever. High-quality data is the foundation that allows organizations to delegate judgment and tasks traditionally handled by humans to AI systems. However, the volume and diversity of data required in real-world environments are enormous. While structured data is relatively easy to process, the true challenge lies in unstructured and semi-structured data, which far exceeds structured data in both scale and complexity. This challenge is further compounded by differences in data storage formats and processing methods across organizations.
Miso Information Technology addresses these challenges through its proprietary platform, SmartBIG, which Nam identified as the company’s core competitive asset. He highlighted two key strengths: fabric architecture and data governance.
“Fabric architecture is based on the idea that not everything needs to be stored in a single place,” Nam explained. “It allows us to utilize distributed data sources as they are.” This approach significantly reduces the cost and time required to collect and integrate massive datasets into a central server. Underpinning this capability is robust data governance.
“When multiple systems or departments use data within an organization, the same term can carry different meanings,” Nam said. “Data governance ensures that a data element has the same definition and standard everywhere, preventing confusion.”
In SmartBIG, only metadata—such as data location and attributes—is centrally managed, while the original data remains in its source systems. This enables users to quickly identify and access the data they need at a specific moment. As a result, fragmented datasets can be utilized as if they were standardized, without being physically consolidated. Nam noted that while fabric architecture is often discussed as a marketing concept, Miso Information Technology has implemented it as a practical, working technology.
Federated Learning Know-How for Sensitive Data Environments
These capabilities are proving especially valuable in the medical and healthcare sectors, where data regulations are among the strictest in the AI market. A notable example is the company’s recent collaboration with Gangbuk Samsung Hospital to develop disease prediction models based on phenome data from 1.5 million individuals.
Phenome data refers to physiological and biological characteristics that emerge through interactions among genetic, environmental, and lifestyle factors. Such information is classified as sensitive personal data and is subject to strict access and usage restrictions. However, to realize truly personalized and precision healthcare in the AI era, these barriers must be addressed without compromising privacy.
Miso Information Technology has tackled this challenge through federated learning.
“Patient data never leaves the hospital,” Nam emphasized. “We only receive AI models that have already been trained.”
Under this approach, each hospital trains its own AI model internally using anonymized patient data. Only the trained models—not the underlying data—are shared with Miso Information Technology, which then integrates them into a single, more powerful composite model. This structure enables the development of AI models optimized for specific diseases and patient groups while fully addressing personal data protection requirements.
Nam added that a well-implemented federated learning framework can be expanded nationwide to hospitals and clinics, as it significantly reduces the burden associated with handling sensitive personal information. Through this approach, Miso Information Technology is positioning itself as a key enabler of trustworthy AI in highly regulated domains.
댓글
댓글 쓰기