Antiviral peptides (AVPs) hold potential as therapeutic agents against viral infections such as Enterovirus 71 (EV71), known for causing hand, foot, and mouth disease. However, the discovery of novel AVPs is hindered by the complex dynamics of peptide-virus interactions and the expansive search space. We explored the application of machine learning techniques to accelerate the discovery of novel AVPs, by training Long Short-Term Memory (LSTM) networks and Variational Autoencoders (VAE) with sequences from AVP databases and generating new sequences that have potential antiviral activities through sampling from the model. By incorporating sequence folding and protein docking, we are able to assess the structural compatibility and binding affinity of these peptides against viral targets. The study demonstrates the capabilities of deep generative models in designing peptides with potential antiviral activities, providing insights into the future development of peptide therapeutics.