Drug discovery projects are optimizing against multiple parameters to achieve their desired efficacy and safety profile. Data can support such optimizations if relationships are extracted as predictions and made available for the team efficiently. This short episode presents predictive model building for passive permeability using publicly available data and the new Trainer Engine from ChemAxon. Feature extraction and results on challenging scaffold-based split scenarios will be highlighted to get to a model that is ready for integration to support medicinal chemistry teams via Design Hub.