Membrane curvature sensing is a fundamental biophysical mechanism 1 that underlies key cellular processes, including vesicular trafficking, membrane remodeling, and the internalization of transmembrane proteins such as G protein–coupled receptors (GPCRs). Beyond their intrinsic biological relevance, curvature-sensing peptides have also shown translational potential, for example in selectively targeting enveloped viral particles 4 or extracellular vesicles in cancer. Despite their significance, relatively few curvature-sensing peptides have been characterized, and a generalizable design framework remains lacking. Amphipathic helices – a class of αα-helical peptides with hydrophobic and hydrophilic faces – have traditionally been considered sensors of positive membrane curvature. We previously showed that they can also recognize negative curvature, broadening their functional scope. Here, we introduce an integrated framework that combines evolutionary optimization with molecular dynamics (Evo-MD) 7 to generate a large large-scale in silico library of curvature-sensing peptides. This approach establishes a systematic platform to probe structure-function relationships and identify new design principles. Guided by these principles, curvature-sensing peptides can be engineered as precision-targeting modules, for example in chimeric constructs with antimicrobial peptides, to achieve controlled delivery of bioactive compounds. By localizing to regions of negative mean curvature, such modules could promote uptake via endocytosis, whereas avoiding these regions may suppress unintended internalization. Our simulations further indicate that curvature-sensing elements can modulate the spatial organization of transmembrane proteins, offering a route to influence membrane-associated signaling. Together, this work presents a computationally driven strategy for the rational design of curvature-sensing peptides, with broad implications for membrane biology, targeted therapeutics, and synthetic biology.