We design Rbx1 binders by first learning broad interface patterns from proteins that naturally engage TfR1 or related surfaces. we define a general signature associated with productive binding at the desired rbx1 N-terminal region. We then generate candidate binders and prioritize those that best match this signature, while preserving sequence and structural diversity. In this way, the approach combines pattern learning with guided design to enrich for candidates that are more likely to bind the target site effectively.