A stent is a medical device used to support and keep open structures such as arteries or airways. Patient-specific stents, i.e., stents that were adapted to the geometry of the specific implantation site, have been shown to be promising [1]. This motivates the investigation of optimization methods for patient-specific stent geometry. While each patient is different, the loss-function and the simulation-model for the optimization are the same, usually to minimize stress concentrations in the vascular system. Thus, finding an optimization method that can exploit these similarities would enable a fast and more efficient design of these stents. Utilizing conventional optimization, each patient-specific geometry would result in a completely new optimization run. While there are methods to speed-up conventional optimization methods, e.g., by utilizing surrogate models, Reinforcement Learning (RL) employs a different approach, where the optimization task is divided into an offline and an online part. During the offline phase, the RL algorithm trains an agent on a dataset of patient-specific geometries. Through trial-and-error interactions, the agent learns a general strategy for optimizing similar patient-specific geometries. During the online phase following a successful training, only a few iterations are sufficient to obtain an optimized design. Due to the additional cost incurred in the offline phase, this approach is particularly attractive if optimal designs need to be generated for multiple geometries so that the offline training is amortized [2, 3]. A similar approach for patient-specific intracranial stenting has already been successfully applied in [4]. In this contribution, we explore different examples towards the goal of optimizing patient-specific stents. We present three examples of increasing complexity, in which an RL agent learns to optimize the geom etry of various slender or fiber-reinforced structures with respect to an external load. These examples explore different aspects of the needed characteristics of a patient-specific stent and are to be understood as the first and necessary step towards our overarching goal.