Simulation is indispensable during the design of many biomedical prostheses that are based on fundamental electrical and electronic actions. However, simulation necessitates the use of adequate models. The main difficulties related to the modeling of such devices are their nonlinearity and dynamic behavior. Here we report the application of recurrent artificial neural networks for modeling of a nonlinear, two-terminal circuit equivalent to a specific implantable hearing device. The method is general in the sense that any nonlinear dynamic two-terminal device or circuit may be modeled in the same way. The model generated was successfully used for simulation and optimization of a driver (operational amplifier)-transducer ensemble. This confirms our claim that in addition to the proper design and optimization of the hearing actuator, optimization in the electronic domain, at the electronic driver circuit-to-actuator interface, should take place in order to achieve best performance of the complete hearing aid.