42Maru Drives AI Innovation for Government, Defense, and Local Authorities

Artificial intelligence (AI) is ushering in revolutionary changes to public service sectors, encompassing administration, national defense, and even the mental health of the populace. Generative AI startup 42Maru is leading the AI transformation (AX) in the public sector by prominently featuring its ultra-lightweight language model (sLLM) and Retrieval-Augmented Generation (RAG) technology. 42Maru participates in the Ministry of Interior and Safety's 'Cross-Government Hyperscale AI Common Infrastructure' project, supporting various government ministries in easily utilizing generative AI. Public officials can delegate repetitive tasks such as administrative document-based Q&A, summarization, and policy draft generation to AI, even without specific development knowledge. This opens an era that maximizes the efficiency of public administration through the shared use of AI infrastructure across ministries. In the defense sector, AI contributes to enhancing combat power. 42Maru, in partnership with the Defense Acquisition Program Administration and the Army Logistics Command, is developing a generative AI-based QA system for the integrated logistics information system. It analyzes vast amounts of maintenance, repair, and legal documents using its defense-specific lightweight LLM and RAG technology. This is expected to automate manual tasks such as automatic searching for maintenance instructions, predicting parts requirements, and summarizing technical data, significantly improving equipment operational efficiency and weapon system sustainment rates. In the future, it plans to evolve into an expandable model encompassing multi-modal data, leading the smart transformation of the logistics support system. AI's touch also extends to national mental healthcare management. 42Maru is accelerating the development of an 'AI-based psychological care counselor assistant service' in collaboration with LG U+, Sungshin Women's University, and Asan Medical Center, among others. The entire counseling process, from collecting client's preliminary information to automating consultation records and crisis monitoring, is supported by the LLM+RAG platform. Notably, a psychological counseling-specific LLM trained on thousands of real counseling sessions will elevate the quality of counseling to another level. It is expected to contribute to strengthening professional counseling capabilities by analyzing clients' lifestyle patterns to identify high-risk groups early, providing emotional status assessment and summary reports before and after counseling, and even embedding counselor training functions. The smart transformation of local government administration is also accelerating. 42Maru, which is jointly pursuing the 'Public Administration Generative AI Convergence Service' project with Busan City, provides an sLLM+RAG-based solution that automates repetitive administrative tasks such as citizen inquiries, policy data search, and press release draft generation. The AI, trained on Busan City's unique administrative data, will serve as a demonstration model that enhances the quality of citizen responses and reduces the workload of public officials. 42Maru mitigates the chronic problem of hallucination in generative AI with its RAG42 technology and AI reading comprehension technology MRC42, and ensures reliability by developing LLM42, a lightweight model optimized for specialized industrial fields. It supports an enterprise private mode, allowing AI to be used safely without concerns about sensitive internal information leakage. Kim Dong-hwan, CEO of 42Maru, emphasized, "The introduction of generative AI in the public sector is more than just a technology transfer; it's about changing the paradigm of how work is done," adding, "We will continue to create innovative public AX cases with domain-specific sovereign AI technology that possesses both accuracy and reliability." Attention is drawn to how 42Maru's technology will transform the future of public services in South Korea.