Computational biomechanical models have been essential in understanding the interplay between neurons, biomechanics, and physical environments. The salamander Pleurodeles waltl is one important model organism for studying neuromechanics behind amphibious locomotion with regeneration capabilities. However, previous salamander models were either strongly simplified in geometry and physics or relied heavily on artificial parameter tuning without an accurate mapping of experimental data. Here, we present a biomechanical modeling and optimization pipeline grounded in animal data to build a biologically plausible biomechanical model for salamander Pleurodeles waltl, which can replicate observed walking and swimming kinematics in a physical simulation. In this pipeline, we model the animal as rigid links connected with passive joints and actuated by antagonistic muscles, while the contact mechanics are modeled by friction and collision between body components and terrain. Model parameters including inertia, friction, and joint passiveness are either measured on euthanized animals directly or on live and anesthetized animals with customized experimental setups. Other parameters that are difficult to measure directly, such as the contact dynamics, the hydrodynamics, and controller gains are initialized with empirical values. To close the reality gap between the measurement and the physical simulator, and to improve the empirically guessed parameters, we optimize the parameter values with an evolutionary algorithm (NGSA-II). The optimized model ultimately generates walking and swimming behaviors corresponding to real animal recordings in the MuJoCo physical simulation environment. Our modeling and optimization pipeline provides a rapid and systematic approach to creating biomechanical models with high representation power grounded in animal data. Moreover, the created models could be further extended with various neural controllers, sensors, and muscle actuators to help study the mechanisms of gait transitions, the interactions between CPGs and sensory feedback, the role of descending pathways in generating rich behavior, and the spinal cord regeneration in a neuroscientific context.