Nota AI debuts Korea’s first VLM-powered real-time surveillance, cutting false alerts and improving complex risk detection.
Nota AI has launched NVA, Korea’s first commercialized Vision-Language Model (VLM)-powered real-time video monitoring solution. Unlike traditional systems that focus on basic object detection, NVA understands the relationships between objects. As well as sequences of events, and operational context. This allows it to detect complex safety violations and multi-factor risks in real time, addressing the long-standing issues of false positives and missed detections.
The solution leverages a natural-language Q&A interface and automated reporting to streamline incident response. In a collaboration with Kolon Benit, NVA proved its capabilities in an eight-scenario proof of concept at Kolon Industries’ Gimcheon 2 Plant. The system identified violations such as “working alone on a ladder without safety gear” and triggered alerts integrated with on-site equipment in under 2.5 seconds.
NVA’s GenAI approach enables rapid rule updates, reducing implementation time for new detection parameters from over three months to less than a week. Full system deployment takes only two to three weeks and integrates with existing CCTV setups, eliminating the need for hardware replacement. This adaptability allows deployment across varied environments, from small sites to large industrial complexes, with security ensured via end-to-end encryption.
Nota AI is now preparing to expand NVA internationally, targeting sectors including industrial safety, transportation, retail, and childcare. The company sees the launch as a strategic milestone ahead of its IPO, showcasing its ability to deliver practical GenAI solutions for high-stakes monitoring. CEO Myungsu Chae emphasized that NVA’s combination of VLM intelligence and operational speed could set a new global standard for industrial AI surveillance.
