We are looking for an intern that will help us to research and create a new architecture to support Edge Machine Learning in our current platform.

For most current Machine Learning use cases models are deployed in a central (cloud) environment. This means that data has to be transferred to that central environment for further processing. With the growth of connected devices (IoT) and increasing amounts of sensor data that needs to be processed, it could be more secure and efficient to do the data processing (in this case machine learning) within the edge nodes themselves. 

In this internship we want to build the first version of our Machine Learning Edge deploy and management feature. This means that we should be able to manage edge machine learning deployments. Making sure to both have an architecture that fits into our current product and is: 

• Efficient: in use of network resources and bandwidth, 

• Accountable: no black-box predictions

• Secure: no concessions on security by moving away machine learning deployments to the edge.

Examples of research questions that come to mind are the following, though we encourage you to introduce your own:

• What would be the best architecture to deploy models from a central platform to an edge device? (e.g. Raspberry Pi).

• What is a generic way to package Machine Learning models for Edge ML deployments? (e.g., Web Assembly, .wasm)