Out of Set Raises Seed Funding to Build On-Device AI Infrastructure and Expand Model Training Capabilities

Out of Set, a South Korean startup specializing in on-device AI, has closed a seed funding round led by The Ventures.

The company develops ultra-lightweight vertical AI models designed to run entirely on-device — without any internet connection. Its technology directly addresses the structural limitations of cloud-based AI, including privacy risks, network latency, high operating costs, and server outages, by processing everything locally on the device itself.

Out of Set was co-founded by two AI researchers with deep domain expertise. CEO Hyungjoo Kim holds a master’s degree in electrical and computer engineering from Seoul National University and spent five years leading voice AI development at NeoSapience and Supertone. An on-device speech synthesis model he released while at Supertone ranked first in its category on Hugging Face and garnered thousands of GitHub stars, establishing his credibility in the global AI community. Co-founder Hyunseung Lee pursued a Ph.D. in speech recognition at Seoul National University and previously designed and built real-time speech recognition systems at AI simultaneous interpretation company XL8.

The company’s initial focus is speech recognition and speech synthesis — two technologies in high demand across security-sensitive sectors such as healthcare, legal, and finance, as well as consumer and industrial applications including smartphones, automobiles, and robotics. Most on-device AI models currently on the market are derived by compressing large cloud-based models, a process that typically takes months before products can ship. Out of Set takes a different approach: its models are designed natively for on-device environments from the ground up, enabling immediate product integration without performance trade-offs.

The seed funding will be used to build model training infrastructure, expand datasets, and hire key talent. The company has also developed a proprietary system for automatically fine-tuning models to meet individual customer requirements — allowing it to deliver customized solutions with fewer resources and in less time than traditional outsourcing approaches.

“Out of Set is a team that can translate research capabilities into real products,” said Cheolwoo Kim, CEO of The Ventures. “With foundational technology that solves the limitations of cloud AI at the device level, we expect them to emerge as a core player in this market.”

“AI performance is determined by data and architectural design, not model size,” said Hyungjoo Kim, CEO of Out of Set. “We’ve designed for real-device deployment from the research stage, and we intend to set the standard for on-device AI.”