Implementation Considerations for Machine Learning at the Edge of the Cloud