The coming decade will see an explosive increase in data generated by edge devices such as connected cars and smart city cameras. Yet 5G network bandwidth will remain a critically limited resource, creating an urgent need for edge devices to determine which data is relevant to send to the cloud for service. Therefore, it is imperative to construct efficient data representations to reduce network traffic by 100x and data storage by 10x. Spero Devices is developing an AI in-memory-computation processor to construct relevant, efficient data representations at 10x greater computational efficiency than the state-of-the-art, enabling low latency at mobile power budget. In addition, Spero is developing an accompanying software platform to seamlessly transform edge data into cloud intelligence, enabling revolutionary new services such as automated selection of AI training data from edge devices.