Physics-informed deep learning has recently emerged as an alternative to  classical solvers for partial differential equations (PDEs). In this  thesis, conditional generative adversarial networks (cGANs) will be used  for simulating seismic wave propagation and seismic  inversion. In the first step, a single cGAN will be trained to learn  wave propagation in an acoustic medium. By incorporating the underlying  physics in the training loss function of both the generator and  discriminator and adding the subsurface velocity distribution as a  condition, we obtain a physics-informed cGAN. In the  second step, a second cGAN will be used in conjunction with the first  one to obtain a subsurface velocity model from measurement data. Using  physics-informed cGANs instead of classical methods should allow for a faster solution of the  inversion problem while still retaining high accuracy. Additionally,  the use of cGANs allows for uncertainty quantification, providing a  possibility of measuring the confidence in the network's prediction.