Advisor Lauren: “Leading AI Transformation in Public, Defense, and Finance with On-Premise AI Agents”

- Advisor Lauren is an on-premise AI solution provider that helps organizations in the defense,public,and finance sectors adopt "AI Transformation" (AX) without the risk of leaking sensitive data to external cloud networks.
- The company differentiates itself through its Forward Deployed Engineer (FDE) system, where teams of AI and business experts work on-site to build custom "Ontologies" and specialized agents for high-stakes tasks, such as detecting defects in wind turbines with 90% accuracy.
- Founded by CEO Min-sung Han, the startup is expanding its reach into Southeast Asia and the Middle East by leveraging high-performance NPU computing and VLM technology to address the skilled labor shortage in critical national industries.


We are living in an era where Artificial Intelligence (AI) is reshaping the industrial landscape. Numerous companies are shouting for AI Transformation (AX) in their own ways, adopting cloud-based solutions. However, there are more than a few organizations that find it difficult to keep up with the pace of change. The defense, public institution, and finance sectors are representative examples. They handle sensitive data that determines national security and the survival of enterprises. Therefore, they cannot store core information on external networks. In this environment, cloud-based AI services are hard to consider as an option from the beginning.

The choice made by the public, defense, and finance industries, which deal with sensitive data, to utilize AI is the construction of internal networks. This corresponds to the On-Premise (internally built) method, where an AI environment is established within one's own data center. The problem is that it is difficult to build AI models within an internal network. This is because there are many variables to consider, such as AI model selection and the scope of data training, and the technical barriers to entry are high.

Min-sung Han, CEO of Advisor Lauren / Source = IT dongA

Is it possible to introduce an AI agent that makes expert-level judgments while fundamentally blocking the risk of data leakage? There is a startup that has suggested a solution to this question. It is Advisor Lauren, an on-premise-based AI solution company that helps the AI transformation of the public, defense, and finance sectors.

Wanted to prepare for the changes in AI after ChatGPT 3.0
Min-sung Han, CEO of Advisor Lauren, built up his experience in AI by conducting machine learning research at the Agency for Defense Development (ADD) and experiencing data-driven decision-making processes at Kiwoom Securities. Believing in the possibility of AI to change the world, he prepared for his startup. The reason for the founding was that he felt a massive flow of change seeing the emergence of ChatGPT 3.0.

He predicted that the era of agents performing tasks on behalf of humans would come naturally. He believed that substantial value is created only when agents that can understand the context of tasks and make judgments on their own are properly established. Naturally, he focused on the on-premise AI infrastructure market, which possesses unique data, rather than general-purpose AI.

“Various agent services are provided, such as writing emails or managing calendars. However, I don’t think those functions have high enough added value to pay a monthly fee. I thought that an agent that moves with a company’s unique data is the truly valuable service.”

Advisor Lauren helps build on-premise-based AI infrastructure / Source = Advisor Lauren

Advisor Lauren starts from the idea that no matter how outstanding an AI model is, it is meaningless if it cannot be used in the field. Therefore, it was judged that simply installing an open-source LLM is not enough. This is because general-purpose AI models are designed for the general public, making it difficult to perfectly perform the unique tasks of a specific company. Taking this as an opportunity, they focused on redesigning core next-generation technologies such as VLM(Vision Language Model), Text2SQL(natural language-based SQL conversion technology), GraphRAG(knowledge graph-based retrieval augmented generation), and Secure MLOps(AI model performance maintenance and management platform) within the on-premise infrastructure to fit the customer’s business structure.

AI that understands and executes in the field
Advisor Lauren’s competitiveness lies in its infrastructure construction and AI integration capabilities. While many AI companies aimed for the Software as a Service (SaaS) model, Advisor Lauren chose vertical expansion. While safely protecting the customer’s unique data, they built an 'Ontology'—a detailed manual that AI can understand by breaking down work units very finely.

A representative example is the wind turbine defect detection project conducted with Korea Southern Power (KOSPO). KOSPO is a public enterprise responsible for 6% to 9% of South Korea’s total power generation, which was separated from Korea Electric Power Corporation (KEPCO) in 2001. Previously, humans took photos of the wind turbine blade defects and judged them with the naked eye, but the problem was the level of proficiency.

To find defects in wind turbines, it is a high-difficulty task that usually requires a senior worker with about 5 to 7 years of experience to show an accuracy of over 90%. If this task is entrusted to general-purpose AI services like ChatGPT or Claude, the accuracy rate is only at the 50% to 60% level. This is because specialized training for the specific purpose was not conducted.

To increase service accuracy, Advisor Lauren directly secured 75,000 photos of defects. Subsequently, they hired 30 personnel, educated them on domain knowledge (Ontology), and labeled the data based on the same criteria. They even passed data quality certification through cross-validation. As a result, they recorded an accuracy rate of 83%, surpassing a first-year worker, and internally completed an advanced agent model approaching 90%. This model obtained a Class A rating (based on over 99.7% accuracy) from Wisestone, an international certification body.

Advisor Lauren focuses on building AI suitable for the customer’s business environment / Source = Advisor Lauren

Han emphasized, "We don't just stop at putting data into the AI model; we take responsibility for the entire process, including analyzing the customer's workflow, planning, refining, and annotating (labeling) the necessary data, followed by training and deployment." This solution, which runs on Korea Southern Power’s internal AI infrastructure, is the result of Advisor Lauren’s tenacity and engineering capabilities.

Advisor Lauren’s efforts do not stop at software. They are evolving into the field of Physical AI, which performs actions in a physical environment. They are developing a system that uses drones to detect defects inside and outside the generator through autonomous flight and performs close-up filming by approaching on its own when an abnormality is discovered.

In addition to the Korea Southern Power case, they worked with Woori Bank to structure financial data based on Ontology. When a user accesses the investment information section in the Woori Bank app, the financial data service provided by Advisor Lauren is displayed.

The structure of helping AI transformation in the public, defense, and finance sectors, which are sensitive to data, is similar to the U.S. data technology company Palantir. Han also explained that the method where field engineers directly learn the customer's business and then convert it into an automated agent structure resembles Palantir.

Advisor Lauren’s AI agents operate centered on the field / Source = Advisor Lauren

From 2026, Advisor Lauren solidified its '2-person 1-team FDE (Forward Deployed Engineer)' system. It is a structure where an AI engineer and a business expert form a team and are deployed to the customer’s site. They do not just sell technology; they participate in the entire process from the stage of defining the customer’s problem to AI design. In essence, it is AI system integration that builds agents tailored to the corporate business environment. This move reflects the corporate philosophy of focusing on the essence of actual use and problem-solving, rather than being buried in prototype demonstrations or technological boasting.

Securing AI talent and entering the global market are challenges to be solved
Although Advisor Lauren has jumped into the rapidly changing AI market, there are still challenges to overcome. Han cited securing high-level AI talent and global market expansion as challenges.

"In the public, defense, and finance industries, just being good at coding isn't enough. We need convergence-type talent who can understand the sensitivity of data and look at the industrial structure and the strictness of security simultaneously. Accelerating entry into the Middle East and Southeast Asia, where demand for AX is rapidly increasing, is also a concern."

Advisor Lauren plans to solve this problem through the influx of overseas talent and technological advancement. First, they signed a business agreement with Universitas Gadjah Mada (UGM) in Indonesia. Through the business agreement, they plan to pre-emptively secure local AI talent and establish a bridgehead for entering the Southeast Asian market.

Technological advancement is being addressed through government support projects. They are improving the technical maturity of VLM and GraphRAG through the hyperscale AI data construction project under the National Information Society Agency (NIA), which is under the Ministry of Science and ICT, and the high-performance computing (NPU) support project of the National IT Industry Promotion Agency (NIPA).

Dreaming of an 'infrastructure' that protects South Korea’s data security and the future of the industry
Advisor Lauren has achieved results such as presenting papers at the Korea Institute of Military Science and Technology, ranking first in the overall score of the NIA hyperscale AI data construction project, winning the Excellence Award in the Korea Credit Information Services D-Testbed, and receiving an excellent evaluation in the NIPA high-performance computing support project. They have proven their capabilities by successfully building a service infrastructure for a large financial platform boasting a monthly active user (MAU) scale of 8 million.

Behind the expansion of Advisor Lauren’s on-premise AI infrastructure business was the support of the Korea University Crimson Startup Support Foundation's Initial Startup Package program. Advisor Lauren received consulting for securing intellectual property rights and global expansion. An opportunity to participate in GITEX, an information and communication technology exhibition held annually in Dubai, was also provided. In this venue, they drew attention by showcasing their drone vision AI model and GraphRAG technology.

Han explained, “For an early startup that must achieve technological development and commercialization simultaneously, the support from the Korea University Crimson Startup Support Foundation, which provided everything from investor networking to participation in global exhibitions, served as nourishment for Advisor Lauren to grow qualitatively and quantitatively.”

Min-sung Han, CEO of Advisor Lauren / Source = IT dongA

"In the U.S., AI is still an option, but in Korea, AI has become a necessity. As the baby boomer generation begins to retire in earnest, a vacuum of skilled personnel is inevitable. Just as India and China rapidly moved to mobile environments, Korea must also hurry its AI transformation to fill the gap of skilled personnel. Advisor Lauren will grow into a company that makes AI work in the deepest and most secret fields in the Korean public, defense, and finance sectors without data leakage."

Advisor Lauren dreams of a world where AI is stably established in the most important fields that the world does not pay attention to. Han believes that the next competitive stage for AI will be the field itself, not the model, and plans to further solidify the structure that integrates the entire process from defining customer problems to AI design and deployment.

By Hyung-seok Kang (redbk@itdonga.com)