Protein engineering is an active area of research in which machine learning has proven quite powerful. Here, the authors present a deep learning method that integrates both general and protein-specific sequence representations to improve the engineering of one’s protein of interest.