In this study, we systematically evaluated four solvent-based pretreatment strategies—including alkanolamines, ionic liquids, organic solvents, and deep eutectic solvents—across eighteen biomass types, including hardwoods, softwoods, grasses, and agricultural residues. High-throughput experiments were conducted under standardized hydrolysis conditions, generating a comprehensive dataset of sugar yields. Alkanolamine-based pretreatment demonstrated the highest average sugar release (85–95%) across diverse feedstocks, highlighting its potential for scalable SAF production. Also, to accelerate optimization and reduce experimental workload, we developed machine learning (ML) models trained on compositional and structural characterization data. Multiple ML algorithms were benchmarked, and the most accurate model was used to predict sugar yields for biomass mixtures, with experimental validation confirming the predictions. Our results demonstrate that ML-guided strategies can effectively streamline biomass pretreatment development, supporting the deployment of flexible, feedstock-agnostic biorefineries.