This study introduces a new methodology for predicting the fatigue behaviour (S-N curve) of Ti10Ta alloys manufactured via Laser Cladding, considering addictive manufacturing defects analysed through nanotomography and optical microscopy. The existing models inadequately address the impact of internal manufacturing defects on fatigue performance, particularly in Ti-Ta alloys obtained by additive manufacturing. Optimisation algorithms Nelder-Mead, SLSQP, and L-BFGS-B were applied to refine the new model parameters and improve prediction accuracy. Experimental data, including intrinsic manufacturing defect characteristics and fatigue tests, were used to evaluate the new models, achieving a high prediction accuracy (R2 = 91.2%) comparable to Basquin’s equation (R2 = 92%). The results revealed that while all algorithms improved initial predictions, Nelder-Mead was the most robust, and SLSQP converged fastest. Three distinct predictive models were developed, tailored to defect data from nanotomography and optical microscopy. These findings underscore the significance of defect size and distribution in fatigue behaviour and provide a practical framework for characterising Ti10Ta alloys, offering enhanced precision and adaptability for biomedical and engineering applications. This methodology represents a significant advancement in understanding and predicting fatigue performance in advanced titanium alloys.