Eastsoft Wins JDC In-House Generative AI Language Model Project

Eastsoft announced on the 21st that it has won JDC's in-house generative AI language model project. JDC stands for the Jeju Free International City Development Center. The core of the project is to establish a generative AI environment on JDC's internal servers.

The contract is tied to JDC's goal of using its own business materials without exposing them to external internet-based services. The project targets an AI platform that complies with security and personal information regulations, and the project period is 6 months from the contract date. The goal is to support the use of internal data in business operations.

Accordingly, Eastsoft plans to deploy a small language model (sLLM) on an internal server at an organization separated from the external network, and provide a work support platform combined with AI agents. The platform is designed to connect business knowledge distributed across multiple systems and documents within the organization, helping employees quickly find the information they need and supporting their decision-making at work.

JDC will establish specialized chatbots for each business area, targeting major functions such as development projects, duty-free operations, budgeting, safety management, contracts, and human resources. The chatbots will be configured to apply language models of different sizes depending on the type of query, with language models of 30B or larger for routine question-and-answer tasks and language models of 120B or larger for complex long-form reasoning.

To ensure response quality and freshness, the company will use retrieval-augmented generation (RAG) technology for document search. It will embed about 3.8 TB of documents inside the organization and deliver answers in a structure that provides evidence and the latest information together by linking existing systems such as the enterprise content management system (ECM) with external data.

JDC will go beyond simple question-and-answer functions and also set up an AI agent environment for actual work execution. To this end, it will apply execution-type agents based on API calls, provide an agent builder that can design and run workflows in which multiple agents collaborate, and support the expansion of the scope of AI applications in the future.

In the operations phase, the company will establish a security and monitoring system tailored to a closed network environment and deploy LLMOps. Eastsoft plans to configure the system with public-sector work characteristics in mind. The system will be set up to present supporting documents for its answers and to enable linkage with real work processing.

The project will apply technology accumulated in Eastsoft's agentic AI service, Alan, and will also draw on AI orchestration capabilities. Eastsoft has recently applied these technologies to agriculture and defense. The JDC project is significant in that it expands the scope of sLLM and AI agent applications to the public sector.

Eastsoft CEO Jung Sang-won said the project is about building an environment where generative AI can be used for internal work even in public institutions where security is critical. He said the company will support JDC's business innovation based on the agentic AI capabilities it has accumulated through Alan. He also said it will pursue the expansion of public AI projects.

Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=408535

References

• https://www.epnc.co.kr/news/articleView.html?idxno=408535