A hybrid cloud-edge infrastructure is particularly advantageous for machine learning/AI applications. These typically involve complex computations such as the synthesis of training data or the optimization of neural networks. Performing these on a central entity (a public or private cloud) and sharing the results with one or multiple consumers has obvious advantages. Inference can be run on every consumer separately and in secure isolation from the training instance in the cloud. 

The API runs on a piece of hardware within an isolated network on the shop floor. All resources necessary for operation (applications, trained AI models, configurations, etc.) must be uploaded to the device prior to deployment or through a VPN tunnel, so that it is able to process all necessary requests made by the external application (e.g., the robot motion program).

When network latency is not of concern, all API functions are initiated by making HTTPs requests directed to one of our cloud endpoints. This way, our software integrates into any application with access to the internet — without the hassle of commissioning dedicated hardware.