Beneath the earth there are many structures, such as different types of rocks and salts. Among them are also hydrocarbons that are a valuable resource for the oil and gas industry. One way of studying sub surfaces is using seismograms, which offers a seismic-wave representation with many valuable information of the area. By studying the patterns within the seismic data one can generate a representation of the subsurface based on some parameters that are able to show each one of underlying structures, such as the velocity that the waves propagated. With the advancement of computer-related technology, such as multi-core processors and GPUs, the processing power of computers have increased and the possibility of working with a much larger amount of data and using new and more powerful computational techniques, such as deep learning, was made possible in a variety of fields. Recently, deep learning methods are being applied to solve many geophysical problems, including the estimation of subsurface structures based on the velocity parameter. This work shows an interdisciplinary approach to estimate velocity models from computer modeling seismograms of non-real sub surfaces using a supervised learning artificial intelligence technique. The results obtained can contribute much to the scientific community as it demonstrates how changes in the seismic data modeling process reflects in the velocity model estimation.