Physics-informed neural networks (PINNs) can be viewed as an alternative to traditional finite element methods (FEM) for solving partial differential equations in the modeling and design of semiconductor devices. Implementation of physical constraints or laws is achieved by adding a proper loss function term to the total loss function. This modular approach makes straightforward the integration of additional physical models. Using as benchmark a simple semiconductor structure, we addressed direct problems by predicting its electrostatic potential, and inverse problems by reconstructing the dopant concentration profile from the electric field. Although some noticeable errors appear, the overall results are consistent with FEM analysis predictions and show the potential of PINNs as inverse design tools that do not pretend to replace traditional methods but rather complement them..