On-device specialist joins state-led push to put Korean AI models on real-world hardware. K-Moonshot targets higher research productivity and lower reliance on foreign AI infrastructure
Nota recently joined Korea’s government-led “K-Moonshot” program, a national plan to apply artificial intelligence to scientific research and industrial innovation. The Seoul-based optimization start-up will focus on making “physical AI” practical by efficiently running Korean AI models on robots, factory systems, and other edge devices.
The K-Moonshot strategy, led by the Ministry of Science and ICT, brings together 33 AI and ICT companies, including LG AI Research, SK Telecom, Naver Cloud, KT, and Upstage, under a single public–private framework. The program aims to double research productivity by 2030 and to tackle eight priority fields and 12 mission-style challenges by 2035, using AI agents, shared computing infrastructure, and standardized “AI-ready” research data. It also seeks closer cooperation on GPU and NPU infrastructure to localize more of Korea’s high-performance computing stack.
Within this ecosystem, Nota will provide its in-house platform, “NetsPresso,” which compresses and optimizes AI models so they can run directly on devices rather than in the cloud. The company says NetsPresso can cut model size by up to 90% while preserving accuracy, enabling deployment in hardware-constrained environments such as industrial robots, factory sensors, and other physical systems. On-device execution is a core requirement for physical AI, where models must respond to the real world with low latency and acceptable power consumption.
Nota also plans to act as a bridge between domestic AI models and a growing range of Korean NPUs. Rather than tuning each model for a single chip, the firm aims to offer a common optimization layer so that foundation and domain-specific models developed under national AI projects can run reliably across different local processors. That approach is intended to support a cycle in which Korean models and hardware are adopted together, lowering infrastructure costs and reducing dependence on foreign cloud and semiconductor vendors.
“Joining this public–private effort to strengthen Korea’s science and technology is highly meaningful,” CEO Chae Myoung-soo said, adding that the company will use its on-device AI expertise to speed up the commercialization of physical AI and deliver practical, optimized models for real research sites. For global investors and industry players, the K-Moonshot partnership suggests that Korea is not only building large-scale AI models but also trying to carve out an edge in hardware-aware, industrial, and edge AI, where physical deployment is becoming the next competitive frontier.
By Min Byung-kwun, KDFN kdf＠kdfnews.com Source_ⓒNota
Tag #노타 #채명수 #문샷
댓글
0