The Korea Artificial Intelligence and Software Industry Association (KOSA) said on the 2nd that the association-run "Korea full-stack AI consortium" completed a proof of concept (PoC) of artificial intelligence (AI) technology based on Aramco Digital's industrial field application cases.
The PoC was pursued under an "AI full-stack cooperation memorandum of understanding (MOU)" that KOSA and the Korea full-stack AI consortium signed with Aramco Digital in Feb. It integrates domestic corporations' AI Semiconductor and cloud, large language model (LLM), and application services into the process of handling complex industrial engineering data.
Seven domestic corporations took part in the PoC, including MegazoneCloud, FuriosaAI, Rebellions, Upstage, LG AI Research Institute, NC AI, and URACLE. MegazoneCloud handled AI and multi-cloud infrastructure buildout and operations, while FuriosaAI and Rebellions were each in charge of AI Semiconductor. Upstage and LG AI Research Institute provided industry-specialized LLMs, NC AI handled 3D modeling, and URACLE provided LLM and Neural Processing Unit (NPU) operations and AI resource management technology.
The PoC proceeded in two phases based on industrial use cases jointly selected with Aramco Digital. In the first phase, it reviewed the applicability of technology that uses AI to interpret 2D engineering drawings and documents and convert them into 3D models, along with integrated control functions.
In the second phase, it verified the process in which AI interprets and structures unstructured drawings and documents and generates them into 3D models. It linked everything into a single architecture, from inference using domestic AI Semiconductor to building a multi-cloud environment and responding to security, and integrated operations of AI services and infrastructure.
KOSA said it confirmed, through this PoC, the potential to apply overseas in industrial sites a full-stack AI model that combines domestic AI Semiconductor and cloud, LLM, operations management, and industrial application services.
KOSA President Cho Jun-hee said, "This PoC is significant in that domestic AI corporations combined their respective specialized technologies to jointly review real demand in overseas industrial fields," and emphasized, "If AI infrastructure and models are overly dependent on a specific external service, operational risk can arise, so from a sovereign AI perspective, it is important to secure in-house AI infrastructure and models, and operations management capabilities."