<p>Road maintenance and rehabilitation is the backbone of economic growth and social connectivity for sustainability and performance. The mechanistic-empirical design methods rely on the material stiffness of pavement layers, back-calculated from the deflection measurements taken from in-service pavements. This study presents a novel approach of using advanced machine learning models to predict pavement layer moduli and evaluate remaining service life (RSL) in terms of fatigue and rutting life. Utilizing field data from various homogeneous sections of flexible pavements including falling weight deflectometer deflections and pavement thickness, algorithms such as random forest (RF), recurrent neural networks with long short-term memory, and extra trees were deployed to model pavement performance. RF emerged as the most robust predictor, achieving an <i>R</i>² of 0.96 and an RMSE of 0.17, demonstrating its ability to capture complex interactions in pavement data accurately. IITPAVE software was employed to quantify critical stresses and strains, further validating the predictive models. The results revealed significant variations in RSL across various sections, with some segments requiring immediate intervention due to elevated strain values, while others exhibited exceptional structural performance. This innovative approach advances the pavement engineering field and provides actionable insights for policymakers and engineers, contributing to optimized maintenance planning and infrastructure longevity.</p>