The application of machine learning in recent decades has generated a digital transformation in different areas of knowledge from medicine to physics, including the social sciences.In geosciences it has been used more and more in recent years as it provides new methodologies based on high-performance computing resources and cloud computing, lowering costs and simplifying prospecting processes.This research aims to explore the application of these new technologies in the reconstruction of well logs.In the present work, four wells located in the Middle Magdalena Valley basin in Colombia are specifically used, wells W1, W2, W3 belonging to the same field and well L located in a neighboring field at approximately 16 km.The methods of Random Forest, Gradient Boosting, and Artificial Neural Networks were used to train models based on the logs of the first three wells that comprise Gamma Ray (GR), Spontaneous Potential (SP), Sonic (DT), and Density (RHO) to later reconstruct the Compressional wave delay time (DTCO) and Shear wave delay time (DTSM) in the neighboring field.We propose a generalized computational methodology to obtain reliable models from known logs to be applied to distant wells located in fields with different structural conditions, but with lithological similarities.For this, the models trained with data from W1, W2, and W3 on L were applied with control logs that allowed a validation of the method and confirmed the possibility of generating the missing logs to improve the well modeling.