Standards-based levee design aims to protect against events with specific probabilities, for example eliminating overtopping from a storm surge with a 1% annual exceedance probability (i.e., a “100-year” event). This allows levee segments to be analyzed independently but ignores interior dynamics and overall risk. We present and implement a framework for calculating optimal risk-informed design heights. Using this design paradigm and multi-objective evolutionary algorithms, we identify levee and floodwall design heights that minimize the total system cost and expected flood losses over 50 years. With our model, decision makers may feasibly evaluate hundreds or thousands of alternative designs over a large ensemble of future states of the world. Comparing to the existing design elevations of the Larose to Golden Meadow Hurricane Protection Project in coastal Louisiana, over multiple climate change scenarios, we identify system configurations of similar cost that reduce the expected value of discounted residual risk of 26%–73% ($8–85 million). We also achieve the same residual risk at 90%–97% of the cost of the existing system (saving $19–73 million).