 Octonion provides a self-care system for industrial IoT products powered by its Edge AI engine to forecast machine health. The specific feature of Octonion's AI is the capability to run on an MCU with no prior knowledge needed. Octonion's partners provide discovery kits running on STM32 for a customer to experiment quickly with the benefits of Octonion's technology. At Octonion we provide a simple indicator, machine health, which helps to anticipate machine degradation. Any customer can start using the machine health as soon as it's deployed. Only a few days is needed for automatic machine health calibration. Then monitor this simple indicator to intervene on the machine at the right time, even before a human inspection can notice the degradation. This is Octonion's key feature for all our predictive maintenance projects. Let's see how this indicator is used by one of our customers. The Willow Group is one of the world's leading premium providers of pumps and pump systems for the building services, water management and industrial sectors. Willow's innovative solutions, smart products and individual services move water in an intelligent, efficient and climate-friendly manner. What is Willow trying to demonstrate by analyzing the pump's behavior at the machine level? The notification of a customer in case of errors in our pump system is of high interest for the building. However, due to the very long lifetimes of our products it is difficult to gather sufficient amount of errors during development or qualification. Regarding this contradiction, Octonion's approach of an anomaly detection seems to match our requirements. Why does Willow choose Octonion's machine intelligence to reach this goal? Octonion came to us with a small demonstrator showing how easy anomalies can be detected by their system. This is of use impressed us. Furthermore, the mature connectivity and club visualization contributed to our decision to water tunnel. For which type of product does Willow force C using this technology? This technology, which we are evaluating, can be used for every product having full parts and shall be monitored. Hence, potentially it can be applied to all our products. What have you validated up to now? What are your next steps? Currently, we are in the validation process trying to prove that all different error states of pumps can be detected by Octonion's technology. We've seen the simplicity of Octonion's machine learning, but monitoring real pumps means adapting the algorithms to real use cases to get the expected output.