Nota AI’s Leap to IPO: Preliminary Fast-Track KOSDAQ Approval

Key Highlights

🔹 Accelerating IPO Momentum with Fast-Track Approval

🔹 Expanding Embedded AI Reach with Renesas

🔹 Redefining Efficient VLMs with Video Self-Distillation

Accelerating IPO Momentum with Fast-Track Approval

We’re thrilled to share that Nota AI has secured preliminary fast-track approval for KOSDAQ listing under Korea’s Technology Special Track. Achieved just two months after filing, this milestone comes in a year marked by intensified regulatory scrutiny and declining approval rates. The approval is more than a gateway to IPO; it’s a clear validation of our technological excellence, commercialization track record, and future scalability. As the first AI optimization company to receive this status, Nota AI stands as a global proof point that efficient, on-device intelligence is not only viable—it’s already reshaping industries.

What Drove the Approval?

• Commercial Viability in AI Optimization: Nota AI’s proprietary NetsPresso® platform is an end-to-end solution for building and deploying lightweight AI models tailored for real-world applications. Trusted by global semiconductor leaders—NVIDIA, Samsung Electronics, Arm, Qualcomm, Sony, and Renesas—NetsPresso® powers scalable AI on compact hardware, enabling fast and efficient deployment across industries.
  ‍
• Strategic Expansion of On-Device AI Across Verticals: Building on this foundation, Nota AI has expanded into key verticals—such as industrial safety, surveillance, and intelligent transportation systems (ITS)—through strategic partnerships and successful commercial deployments. Among the solutions supporting these initiatives is the Nota Vision Agent (NVA), an advanced video monitoring system powered by the company's proprietary Vision Language Model (VLM).

What’s Next in the IPO Journey?

This IPO momentum reflects not just Nota AI’s readiness, but the market’s trust in scalable, real-world AI that works—now and at the edge. Nota AI aims to complete its IPO in the second half of 2025, and we look forward to accelerating this momentum by continuing to build trust with global partners, deliver value to customers, and lead innovation in on-device AI.

👉 Read the full announcement

Expanding Embedded AI Reach with Renesas

Nota AI has partnered with Renesas Electronics, a global leader in advanced semiconductor solutions, to co-develop an optimized Driver Monitoring System (DMS) for Renesas’ newly launched RA8P1 Microcontroller Unit (MCU). This collaboration reflects a shared vision to make efficient, real-time AI accessible across size- and power-constrained embedded environments. By enabling AI to run directly on compact MCUs, the joint solution brings intelligent driver behavior analysis to vehicles that require high accuracy within limited processing resources.

What Makes This Joint Solution Stand Out?

Nota AI’s DMS solution achieves approximately 50 frames per second (FPS) on the RA8P1—demonstrating industry-leading performance for real-time in-vehicle monitoring.

Nota AI has successfully deployed use cases like People Detection, Line Crossing, and Crowd Analytics across Renesas hardware, showcasing consistent performance and deployment readiness. This ongoing partnership reinforces both companies’ commitment to expanding AI deployment across embedded platforms—from entry-level MCUs to advanced MPUs.

👉 Learn more on Renesas Partner Page

👉 Read the full PR announcement

Redefining Efficient VLMs with Video Self-Distillation

We’re excited to share that Nota AI’s latest research has been accepted at the ICML 2025 Workshop, one of the world’s premier machine learning conferences.

Titled "Video Self-Distillation for Single-Image Encoders: Learning Temporal Priors from Unlabeled Video," the study tackles a core challenge in Vision-Language Models (VLMs): their inability to interpret temporal cues from static images.

Why This Research Matters

• Overcomes a Core Limitation of VLMs: Conventional VLMs rely on static images and struggle to understand motion. Nota AI’s approach enables single-image encoders to capture temporal context—such as motion and spatial structure—without modifying their architecture.

• Enables Self-Supervised Learning from Unlabeled Video: By using predictive learning, the model learns from raw video data without requiring manual labels or additional sensors—making it scalable and cost-efficient.

• Drives Real-World Impact Across Domains: This breakthrough lays the foundation for smarter, more responsive AI systems in areas such as industrial safety, surveillance, and Vision-Language-Action (VLA) applications.

By bridging the gap between still images and dynamic understanding, this technique brings AI one step closer to perceiving and responding to the world in motion—opening new possibilities for real-world deployment.

👉 Explore our research on the Tech Blog