Eastsoft Starts Building JDC In-House Generative AI System Using sLLM and AI Agents
IT DAILY ·
✦ AI Summary
Eastsoft has won JDC's "JDC In-House Generative AI Language Model (sLLM) Build Project."
The project will be built on JDC's internal servers, and the contract period is six months from the date of signing.
Eastsoft plans to directly build the sLLM in an internal environment separated from external networks and provide a work support platform that combines AI agents.
Eastsoft has begun building an AI platform for public institutions that can use generative AI in an internal environment separated from external networks.
On the 21st, Eastsoft said it had won the contract for JDC's "JDC In-House Generative AI Language Model (sLLM) Build Project." The project will be built on JDC's internal servers, and the contract period is six months from the date of signing.
The core of the project is to establish an AI platform that can safely use data in an internal environment separated from the outside. Eastsoft plans to directly build the sLLM on the agency's internal servers, isolated from external networks, and provide a work support platform that combines AI agents on top of it.
Through this, Eastsoft plans to connect dispersed work knowledge within the agency and help employees quickly and accurately search for the information they need. It also plans to support evidence-based decision-making, improving employees' work accuracy and speeding up task processing.
JDC is pursuing the build-out of specialized work support chatbots for major business areas and aims to create an answer system suited to each task type by applying language models of different sizes to chatbots divided by business field. Target areas include development projects, duty-free business, budgets, safety management, contracts, and human resources. For everyday Q&A, it plans to apply a language model of 30B or larger, and for complex long-form reasoning, a language model of 120B or larger, to deliver optimal response quality for each type of work.
JDC will embed about 3.8TB of in-house documents and provide a RAG-based AI document search service. It also plans to link the system with existing systems, including the ECM, as well as external data to keep search results and answers up to date.
JDC is also setting up an AI agent environment that can carry out actual work, going beyond simple Q&A. To that end, it will build execution-type agents based on API calls and provide an agent builder that allows collaborative workflows involving multiple agents to be designed and executed directly. Through this, it also aims to lay the groundwork for expanding JDC's own range of AI use in the future.
The project also includes building a security and monitoring system tailored to a closed network environment and establishing LLMOps, through which it will secure a stable and sustainable operating framework.
Eastsoft CEO Yang Seung-gap said the project will serve as a safe and practical generative AI use case for a public institution that prioritizes security, adding that the company will support JDC's business innovation with Allen-based agentic AI capabilities and pursue a rapid expansion of its foothold in the public AI market.
Source: IT DAILY · Yang Seung-gap
Original: https://www.itdaily.kr/news/articleView.html?idxno=242343
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.