Structural proteins enable spectacular mechanical feats in biology, including the persistent attachment of mussels to all underwater surfaces or the ability of the locusts to jump many times their height. At the core of this performance is the sequence of amino acids, each of which has evolved over billions of years to achieve its desired function. However, it is still impossible to predict and engineer the function of proteins directly from their amino acid sequence. This research imitates the natural evolutionary processes to develop robust biomaterials and adapts them to (1) develop biomaterials with unprecedented performance and (2) discover how key aspects of their amino acid sequences lead to this performance. The proposed “directed evolution” approach combines techniques in synthetic biologywith novel methods for measuring the mechanical properties of thousands of samples. In the proposed data-driven approach, these experimental results over large datasets are iteratively informed by simulations and machine-learning algorithms.