TraceWeave has raised approximately $2.26 million (KRW 3 billion) in seed funding to advance its on-device artificial intelligence architecture, with NAVER D2SF joining the round alongside lead investor Mirae Asset Capital.
The company was valued at approximately $22.61 million (KRW 30 billion) on a pre-money basis. TraceWeave plans to use the financing to expand its research and engineering team, continue developing next-generation AI architectures, and grow commercialization efforts across mobility, robotics, drones, smart glasses, wearables, and other device categories.
TraceWeave is developing Phase-TITAN, a post-Transformer AI architecture designed to support long-context reasoning and multi-step agent workflows while keeping memory usage constant as context length increases.
Conventional Transformer models rely on key-value cache, or KV-Cache, which grows as context length expands and can create significant memory requirements. That presents challenges for deploying AI agents on devices with limited memory, battery capacity, and thermal headroom.
TraceWeave said Phase-TITAN eliminates that scaling bottleneck by replacing the conventional Transformer architecture rather than modifying existing open-source models. According to the company, at a one-million-token context without quantization, a comparable Transformer’s KV-Cache can consume tens of gigabytes or more, while its one-billion-parameter Phase-TITAN model uses approximately 0.04 GB.
The architecture is intended to let AI agents call tools, interpret results, and determine subsequent actions entirely on-device without relying on cloud calls. TraceWeave said internal benchmarks using RULER and BFCL indicate Phase-TITAN can maintain performance comparable with Transformer models of the same size and larger.
The company has already demonstrated a multi-step AI agent running on a smartphone using only the device’s CPU and without cloud access.
TraceWeave has also completed a paid pilot with the autonomous driving affiliate of a global automotive group. In that project, its system identified driving-data frames intended to improve the partner’s autonomous driving model without human annotation. The company is additionally validating in-vehicle agent performance with a global automotive electronics software company.
Beyond automotive use cases, TraceWeave is exploring deployments across robots, drones, smart glasses, and wearable devices. The company was also selected as a Top 20 finalist in NVIDIA’s 2026 Inception Grand Challenge.
TraceWeave was founded in October 2025 by researchers with backgrounds at ETH Zurich, Google, and KAIST. The team combines expertise in computational neuroscience, physics, AI architecture, and business development and is based at NAVER D2SF @Gangnam.
NAVER D2SF said the investment aligns with its broader focus on on-device AI and next-generation computing technologies that can reduce reliance on cloud infrastructure while supporting privacy and lower operating costs.
KEY QUOTES:
“Taking the KV cache out was never the hard part. Staying capable without it was. The best part is no part. Phase-TITAN already delivers Transformer-class capability without the Transformer’s memory tax, and the next one goes well beyond its class, with the memory still fixed.”
Jaeyoung Park, Co-Founder And CEO Of TraceWeave
“On-device AI is a core technology for implementing complex, advanced AI across diverse device environments beyond PCs, and it also offers advantages in reducing cloud call costs and protecting privacy. TraceWeave is a team that is innovating AI architecture to overcome the limitations of existing AI models and bring AI to on-device environments.”
Sanghwan Yang, Head Of NAVER D2SF
