A wide range of parameters impact the success or quality of a laser process. Currently, in Laser welding, the first step in the search for optimal parameters is finding the ones that give the weld-penetration needed. This can be done by trial and error, or better yet, using a plan of Experiment. Both of which are costly and time-consuming methods. Meanwhile, Artificial Intelligence offers a new way to estimate parameters. Different AI models were trained on a dataset composed of weld-height measured for 390 sets of parameters across 5 different materials, then compared. A Shallow Neural Network is found to have the best performance, with a R² of 0.95 and a Mean Squared Error of .05 mm² using the usual evaluation method. Notably, on a challenging out-of-training evaluation like predicting penetration on Copper, a Gaussian Process Regressor manages to keep an acceptable R² of 0.72 and a MSE of .24 mm². Using this model the other way around, one can get a set of process parameters for a given weld height and material.