On June 26, 2025, Atomrock CEO Kevin Tseng presented at the AI Taiwan Future Commerce Expo on the topic “When Machines Start to Speak — How Audio AI Can Predict Manufacturing Risks.”
Kevin shared a case from a tire roller system, where each tire produces a distinct sound when passing through the system. When the tire wears, the sound changes. Factories had historically replaced all tires every six months as a precaution — an approach that ignored actual usage differences between vehicles and led to unnecessary labor and material costs.
With audio recognition models deployed on edge devices, Atomrock now identifies when a tire needs replacement based on the sound it emits — shifting from time-based maintenance to condition-based decisions. These models are trained not on public datasets but on sound collected at the customer's own factory, processed in Atomrock's cloud, and deployed back to the edge for real-time inference. Consistent data source leads to reliable performance.
Beyond audio, Atomrock works with visual and sensor data, integrating enterprise data with on-site devices to create AI that fits the environment and runs in production — already deployed in smart cities and port logistics in Japan.
