Naver D2SF Invests in Traceweave

Naver’s startup investment group D2SF said on the 21st that it had invested in Traceweave, an on-device AI startup. It cited the company’s technology for running AI agents on devices with limited hardware resources, including robots and cars. Founded in October last year, Traceweave raised about 3 billion won in its seed round at a valuation of about 30 billion won. Mirae Asset Capital led the investment.

Traceweave developed an AI architecture called Phase-Titan to address memory bottlenecks (limits on memory processing) in on-device environments. The company says the technology keeps memory usage steady even as the amount of information being processed grows. On devices with limited resources, such as robots, cars, smart glasses, and drones, it can support AI agents that understand long contexts and carry out multistep tasks.

Traceweave’s technology supports agents that call external tools on the device without sending requests to the cloud, interpret the results, and decide what to do next. Naver D2SF said it invested because the technology can run agents on limited hardware as AI use expands across different devices.

Yang Sang-hwan, head of Naver D2SF, said Traceweave improves AI architecture to address the limitations of existing AI models and enable on-device AI. He said Naver D2SF has identified technologies for emerging device environments, including AI optimization and compression company Nota, and plans to continue investing actively.

Naver’s startup investment group D2SF said on the 21st that it had invested in Traceweave, an on-device AI startup. It cited the company’s technology for running AI agents on devices with limited hardware resources, including robots and cars. Founded in October last year, Traceweave raised about 3 billion won in its seed round at a valuation of about 30 billion won. Mirae Asset Capital led the investment.

Traceweave developed an AI architecture called Phase-Titan to address memory bottlenecks (limits on memory processing) in on-device environments. The company says the technology keeps memory usage steady even as the amount of information being processed grows. On devices with limited resources, such as robots, cars, smart glasses, and drones, it can support AI agents that understand long contexts and carry out multistep tasks.

Traceweave’s technology supports agents that call external tools on the device without sending requests to the cloud, interpret the results, and decide what to do next. Naver D2SF said it invested because the technology can run agents on limited hardware as AI use expands across different devices.

Yang Sang-hwan, head of Naver D2SF, said Traceweave improves AI architecture to address the limitations of existing AI models and enable on-device AI. He said Naver D2SF has identified technologies for emerging device environments, including AI optimization and compression company Nota, and plans to continue investing actively.