<p>The interface of biomolecular condensates plays a key role in processes such as protein aggregation and biochemical reactions, making it an attractive target for engineering condensates. However, the molecular grammar that drives preferential localization to condensate interfaces remains poorly understood. Here, we develop a computational pipeline that integrates high-throughput coarse-grained simulations, machine learning, and mixed-integer linear programming to design peptides that partition at the interfaces of defined condensate targets. We validated the workflow by designing and synthesizing peptides that localize at the interface of three distinct condensates formed by intrinsically disordered protein regions. These peptides exhibit surfactant-like architectures. Specifically, one tail inserts into the condensate and is enriched in aromatic residues, while the opposite tail is excluded from the dense phase, with its sequence varying according to the scaffold’s net charge. Overall, our pipeline offers a general strategy for rationally designing interface-localizing peptides and for unraveling the governing design principles.</p>