According to an aspect, a method includes receiving a first three-dimensional model (3D) model of at least a body part of a person, receiving a second 3D model of a wearable device, and predicting, by at least one machine-learning (ML) model, a plurality of contact points between the first 3D model and the second 3D model.