||Process design and optimization are challenging task not only because of the model formulation and expensive computation but also numbers of physicochemical parameters deducing from experimental data. Numbers of process design employing novel solvents and producing uncommon chemical, therefore, have been suffered from unknown physicochemical properties and resulting process models inherently has high degree of uncertainty. In this work, we developed and assessed a machine learning methodology to estimate parameter uncertainties, specify solvent physicochemical properties, and evaluate the reaction kinetics of a water-lean amine solvent for a CO2 capture process. We integrated two fundamental methodologies to decrease the experimental and computational costs. Gaussian process Bayesian optimization was applied to the pilot-scale tests; in addition, a rigorous process model employing a newly proposed hybrid Bayesian inference was used, which reduces the computational time of sampling. The assessment highlights the Gibbs free energy of the particular electrolyte as the most sensitive parameter to match the process responses. Both water and water lean amine solvent, K(2)Sol, were observed to act as dominant bases in the absorption kinetics. Furthermore, most output responses of the process model were located in the 95% confidence interval. Our methodology efficiently incorporates process optimization from past experiments and simultaneously identifies solvent characteristics to build rigorous process models that automatically consider uncertainties.