Motivated by the fact that the interaction between Nipah virus F protein and ephrin-2B receptor is amongst the highest known for any viral protein (reference: https://doi.org/10.1016/j.cell.2025.02.030), this system is taken as reference. First, to sample stable proteins designs, the CATH database is parsed with a sequence length limit of 250. Then, TM-align is used to determine which folds are similar to that of the wild-type binder (ephrin-2B) when bound to Nipah's F-protein. Folds with good scorings, specially focused on the binding region, are selected. The binding interaction between those selected domains and the receptor are modeled with Boltz-2, and those with a good scoring are selected, and the binder is re-designed with proteinMPNN, with SolubleMPNN weights at different sampling temperatures for proper sequence diversity. Sequences are then clustered with MMSeqs2 and validated with Boltz-2