Arterial stiffness is a key marker of vascular aging and cardiovascular risk, commonly assessed through aortic Pulse Wave Velocity (aPWV). While oscillometric devices offer a non-invasive alternative for measuring aPWV, their cost and maintenance limit their accessibility in primary care. This study proposes a machine learning model to estimate aPWV using routine clinical variables. A dataset of 390 individuals from the EVASCU study was analyzed, incorporating demographic, anthropometric, and hemodynamic parameters. Feature selection was conducted using Random Forest importance scores and permutation importance. Five predictive models were evaluated via 10-fold cross-validation, with Quadratic Regression ( \(\text {R}^2 = 0.951\) ) and Random Forest (( \(\text {R}^{2} = 0.933\) ) outperforming linear methods, highlighting the importance of capturing nonlinear vascular interactions. Errors were higher at elevated aPWV values, suggesting the need for more data from individuals with vascular impairment. These findings support the feasibility of a cost-effective, accessible Artificial Intelligence-based tool for vascular risk assessment, enhancing early detection and prevention in primary care.