A prominent Fortune 500 insurance company required a way to better predict and plan for abrupt weather disasters. Doing so would help increase reliability of risk models and allow better premiums for policy holders. Knowing that a major weather event had occurred or was about to occur and the capacity to act quickly was important.

Utilizing AWS IoT and Machine Learning services, we created an AI/ML edge solution capable of locally identifying a specific weather event and notifying our client on occurrence. Based on comparable in-market solutions and weather ground stations, utilizing audio as the input signal is a unique approach that the team was tasked to validate for technical feasibility and accuracy. The initial model was built with 10,000 seed audio files spanning various weather conditions.