Keynote: The model efficiency pipeline, enabling deep learning inference at the edge