A computational screening strategy applied to 8 million synthetically diverse ionic liquids, demonstrates its value for sustainable solvent design. In contrast to previous studies which have largely focused on single characteristics, important properties such as viscosity, density, thermal stability and toxicity, were estimated using machine learning. Experimental characterization of 15 compounds selected from the library reinforced the utility of the approach. Over 2600 ionic liquids, a majority of which have never been reported before, were investigated for applications in CO2 and H2S gas separation, and cellulose dissolution, using density functional theory calculations. Among these, around 250 low cytotoxicity and low viscosity ionic liquids merit further investigation. © 2019 The Royal Society of Chemistry.