Rehabilitation robots provide many theoretical benefits to augment the role of a physical therapist; however, to date, therapeutic outcomes following stroke and spinal cord injury have not been improved with the use of rehabilitation robots. Personalized neuromusculoskeletal models have been developed to model dynamic motion and control of the human body, and the state-of-the-art models are capable of including impairment in the model. Incorporating a dynamic model of a rehabilitation robot working in concert with the human limb would enhance the impact of such models in designing personalized treatments. To realize this, the dynamic model of the robot must be solvable in real-time. These combined models can then be used to create personalized, model-based control strategies with the goal of improving therapeutic outcomes through higher subject engagement following spinal cord injury or stroke. To address this need, this thesis describes the design of the MAHI Open Exoskeleton (MOE), a four degree of freedom, serial exoskeleton device for the upper-limb. A dynamic model of the MAHI Open Exo is presented, along with the characterization and friction modeling of the device. The dynamic model provides the basis for a future human-robot combined model, which will be used for personalized, model- based control strategies.