FortyTwoMaru joins Busan's second-phase AI admin project
The AI firm will supply a Busan-specific sLLM and deepen training on administrative data as part of a consortium with the lead contractor and a partner company.
FortyTwoMaru said on 23 October 2026 that it will take part in Busan Metropolitan City's project to expand its AI administrative service, developing core AI models as a member of a consortium.
The project is a second-phase effort to upgrade the functions of Busan's administrative AI platform, 'AI부기주무관'. Under the plan, the city will build a system that operates multiple language models by task and a RAG environment tailored to departments and individual users, and will add AI-based data collection and joint document editing features.
FortyTwoMaru is developing the core AI models together with the consortium's lead contractor, 클라이온, and fellow member 어기야팩토리, according to digitaltoday.co.kr. The company's role covers supplying a Busan-specific sLLM, deeper training on administrative data and optimization of multi-model operation, and it will pursue a pilot application of a pan-government hyperscale AI common foundation.
Busan introduced 'AI부기주무관' last year. The platform is based on an sLLM trained on about 160,000 pieces of city data and administrative documents plus RAG technology, and supports drafting administrative documents, searching ordinances and budgets, and conversational Q&A.
FortyTwoMaru also took part in the first phase of the project. Its CEO said the company will continue working with Busan in the second-phase expansion.
Quick answers
What is Busan's 'AI부기주무관'?
It is Busan Metropolitan City's administrative AI platform, introduced last year. It is based on an sLLM trained on about 160,000 pieces of city data and administrative documents plus RAG technology, and supports drafting documents, searching ordinances and budgets, and conversational Q&A.
What is FortyTwoMaru's role in the second phase?
It is developing core AI models in a consortium with the lead contractor 클라이온 and 어기야팩토리, and will supply a Busan-specific sLLM, do deeper training on administrative data and optimize multi-model operation.