Modern motorsport requires increasingly sophisticated telemetry analysis to extract value from on-board acquired data. One of the primary challenges for track engineers is distinguishing between the natural variability of driver performance and the emergence of actual mechanical or instrumental anomalies. In this context, this thesis work develops and explores the possibilities of a Virtual Sensing system based on Machine Learning algorithms for the dynamic estimation of a racing motorcycle’s response through the prediction of the sensors involved. The activity started with an in-depth literature review, conducted to outline the state of the art regarding telemetry analysis in the Motorsport sector. After, a Data Engineering pipeline was implemented to pre-process the dataset.