Efficient Architecture Paradigm for Deep Learning Inference as a Service | IEEE Conference Publication | IEEE Xplore