fastsolv is a solid solubility predictor that uses an ensemble of fastprop-based neural networks with Sobolev training to predict the logarithmic solubility (logS) of a solute in a solvent at a given temperature, along with aleatoric uncertainty estimates.