The hydroformylation of alkenes is a cornerstone transformation for the chemical industry, central for both functionalizing and extending the carbon backbone of an alkene. In our study, we explored silica-supported crystalline rhodium sulfide nanoparticles as heterogeneous catalysts in hydroformylation reactions, and found that RhxSy systems (x=17, y=15 or x=2, y=3 with 1 wt.% Rh on SiO2) greatly outperform metallic Rh nanoparticles. These systems proved to be exceptionally competitive when benchmarked against other cutting-edge catalysts in terms of activity, with Rh17S15/SiO2 being the superior catalyst candidate. By employing local environment descriptors, unsupervised machine learning and density functional theory, we have examined structure-performance relationships. Our results highlight that the presence of S in close proximity to the catalytic site unlocks the tunability of the surface catalytic properties. This allows for the substrate affinity to be modulated, in particular for Rh17S15, with adsorption energies rivaling those of pristine Rh and improved spatial resolution.