Tailored fibre placement (TFP) is a manufacturing process for composite materials. Continuous fibre rovings are positioned in a textile base material, fixed by a stitching yarn, and are later impregnated in a matrix. Its main advantage relative to classical manufacturing methods is the possibility of manufacturing variable-axial reinforcements with a small radius of curvature. Given the strong dependence on composite material properties regarding fibre orientation, the design freedom provided by TFP can be applied to enhance the mechanical behaviour of many composite structural components. However, when dealing with small dimensions, deviations can be observed between the target path (designed path) and the actual path, effectively stitched by the TFP machine. Among the sources for these deviations, one can mention fibre’s waviness, buckling, friction between materials involved, the influence of input parameters, i.e., the tension applied at the roving, the distance between stitching points, speed, and other systematic deviations inherent to any manufacturing process. Deviations are frequently observed in small radii curves, where the roving shifts the stitching points defining the inner radius. In this context, the present work proposes a machine-learning algorithm capable of minimizing the differences between the target and actual paths by predicting the resulting deviations before the stitching process. The path designed for this training includes straight lines and curves with different radii distributed in mixed sequences and overlapping rovings with and without curvature. Optical measurements are performed after manufacturing the training sample to evaluate the exact roving position and height profile. The algorithm computes the magnitude and direction of the deviations, point by point, and applies them as input parameters in the training process. This hereby calculated output path is a new path that leads to smaller deviations from the target path while placing the roving onto the base material. The production of accurate paths, with minimized deviations, allows further enhancements in the mechanical properties of manufactured composite materials components.