Demands for healthcare tools providing better detections, diagnoses, and treatments are at an all-time high as we gradually transitioning into an aging society. To tackle this challenge, bioelectronic devices made of conformable, soft and biocompatible materials ranging from sensors, actuators, drug delivery devices, to individual computing units were developed to help push forward the healthcare technology. Various translational researches and experimental works have brought forth breakthroughs and innovations from laboratories into clinics, benefitting the quality of life for countless patients. However, the level of understanding in how these devices interact with biology are lacking. As the complexities and types of bioelectronic devices increase, this gap of knowledge makes subsequent device design and improvement difficult, inefficient, and often times sub-optimal. Computational models are ideal candidates to provide insights and guidelines for re- searchers to design bioelectronic devices and better understand their working principles. A proper model not only captures the important components of a system, providing a clear direction for experimental efforts, but also allows the possibility to perform large quantities of in silico studies in a relatively short amount of time. These advantages can reveal a clear relationship of how different components and their properties affect the overall device performance and how bioelectronics interact with biology. In this thesis, two types of neuromodulation devices - the electrophoretic drug delivery device and the spinal cord stimulator were investigated. The working principle of these devices were identified and represented as governing equations. Subsequently, strategies to optimize the performance of existing neuromodulation devices for different applications, and/or new device operation modalities to overcome current limitations due to device archi- tectures were presented. Finally, the theoretical predictions and numerical calculations in this work were validated with experimental measurements. The findings from this work can help design next-generation neuromodulation devices with better efficiency, higher reliability, and greater level of safety for the patients.