Computational peptide design has the potential to significantly accelerate the discovery of novel peptide inhibitors. Forcefield-based approaches have had limited success, and there is increasing interest in the use of machine learning to design peptides and proteins. In this talk, I will discuss how learning the evolutionary trajectories of proteins and peptides can dramatically increase the accuracy of protein/peptide language models. As an example, I will discuss work on spider venom peptides that inhibit human sodium channels. Finally, I will discuss how protein sequence representative embedding models can be used to better analyse deep sequencing data from high throughput peptide and protein screens.