This study aims to assist elders with chronic diseases living in healthcare centers or a big family. In the era of AI, designers are not only to design product forms and functions, but the computer vision technique of AI can be used to help them. Based on the convolutional neural network (CNN) technology (as a supervised learning), three different types of optimizers of MobileNet V2 for training the best model accuracy, which is applied to a gerontechnological pillbox device with face recognition. In addition, the MobileNet V2 model proposed is imported into the JETSON edge machine that can identify different faces and then emit different sounds with its corresponding name and the quantity of medications. The smart pillbox device can improve the compliance and safety of elderly people when taking medicine.