With global push of hydrogen economy, efficient scenarios for hydrogen storage, transportation, and generation are indispensable. Here we devise a strategy for controllable hydrogen fuel storage and retrieval via light-switched CO2-to-HCOOH interconversion. To realize it, palladium sulfide nanocluster catalysts with multiple specific functionalities are directly searched by our home-developed inverse design approach based on genetic algorithm (IDOGA) and ab initio calculations. Over 500 low-energy PdxSy (x + y ≤ 30) clusters are sieved through a multiobjective function combining stability, activity, optical absorption, and reduction capability of photocarriers. The structure-property relationships and key factors governing the trade-off among these stringent criteria are disclosed. Finally, 14 candidate PdxSy clusters with proper sulfidation degree and high stability in an aqueous environment have been screened. Our IDOGA program provides a general approach for inverse search of nanoclusters with any designated elemental compositions and functionalities for any device applications.