Implementing Transfer Learning to Optimize an Edge AI Device