#South Korea #Seoul #AI Startup #NAVER D2SF #TraceWeave

NAVER D2SF Makes Significant Investment in TraceWeave

NAVER D2SF, the venture capital arm of NAVER, has announced a new investment in TraceWeave, a groundbreaking startup specializing in on-device AI. This investment, part of a round led by Mirae Asset Capital, is poised to elevate the tech landscape by advancing artificial intelligence capabilities in devices with limited resources. The team is spearheaded by Jaeyoung Park, co-founder and CEO, who emphasizes the innovative potential of their technology.

TraceWeave is known for its development of Phase-TITAN, a post-Transformer architecture that enhances the ability of AI agents to operate efficiently on memory-constrained devices. Traditional AI models, particularly those based on the Transformer architecture, typically face challenges managing memory usage as context length increases. This limitation has hindered the deployment of advanced AI functionalities across a spectrum of devices, from smartphones to drones.

The unique selling point of TraceWeave lies in its ability to maintain constant memory usage even as the context grows, contrasting significantly with the expansive memory demands of conventional Transformer models. For instance, while a comparable Transformer might require thirty gigabytes or more of memory to handle a million-token context, TraceWeave's Phase-TITAN can achieve similar tasks with only 0.04 GB. This remarkable efficiency opens the doors for AI applications that can operate independently of cloud resources, thereby enhancing privacy and reducing costs associated with cloud calls.

In a recent pilot project, TraceWeave successfully demonstrated its technology's potential in the fields of mobility and autonomous driving. The startup collaborated with an autonomous driving affiliate of a global automotive group, showing that their AI could identify driving data frames without human input, significantly bolstering the partner's autonomous driving model. They are currently validating the performance of their in-vehicle AI agents with major automotive electronics software companies, indicating broader applications of their technology across various sectors.

The founders of TraceWeave bring robust expertise to the table, with backgrounds in esteemed institutions such as ETH Zurich, Google, and KAIST. This mix of engineering talent and business acumen fosters the development of revolutionary technologies tailored to meet the real-world needs of consumer devices across diverse environments.

In light of its innovative potential, TraceWeave has gained recognition, placing as a Top 20 finalist in NVIDIA's 2026 Inception Grand Challenge. The funds from the recent investment round will bolster the organization's efforts to expand its research team and accelerate the development of next-generation AI architectures.

Jaeyoung Park encapsulates the vision accurately, stating, "Taking the KV cache out was never the hard part. Staying capable without it was. Phase-TITAN already delivers Transformer-class capability without the Transformer's memory tax."

Sanghwan Yang, the Head of NAVER D2SF, concurs, noting that on-device AI is crucial for advancing complex AI technologies across a range of devices, while also minimizing cloud dependency and enhancing data privacy. The investment in TraceWeave represents a strategic move not just for NAVER D2SF but for the broader tech industry as it embraces the future of AI technology.