Drag reduction is vital for any new airplane design as the demand for a greener aviation is increasing. Wings and nacelles with laminar flow can reduce the total drag significantly and the backbone for designing laminar surfaces is the transition prediction. The highly complex physics of boundary-layer transition has led to the development of a wide range of prediction methods. Considering their good compromise between accuracy and computational cost, the standard prediction methods used by industry are based on the Linear Stability Theory (LST) and on the Parabolized Stability Equations (PSE). Although these methods have been successfully applied, it can be difficult to make them fully automated due to lack of robustness. Besides, tuning the model setups can be time-consuming. The inclusion of such methods in the aerodynamic design significantly increases the computational cost, especially in optimization loops. The present work proposes a solution for these disadvantages of LST and PSE using a metamodel based on deep learning. Complex Neural networks created using artificial intelligence concepts allows classification and regression of the large datasets necessary for transition analysis. The metamodel reproduces a local stability code and the results are promising both in terms of accuracy and processing speed.