This thesis focuses on advancing the state of the art for calculating the shutdown dose rate (SDR) in accelerator-driven systems, where high-energy neutrons produced by particle interactions activate materials and produce unstable radionuclides. These radionuclides emit photons as they decay, which can pose a hazard to maintenance personnel. Therefore, it's important to compute the SDR as a function of time and space after system shutdown. The SDR can be calculated using the Rigorous Two-Step (R2S) method, which requires a separate neutron and photon transport calculations coupled with an activation calculation. The current R2S workflow for accelerator-driven system relies on physics models in the Monte Carlo N-Particle (MCNP) code to simulate interactions beyond the nuclear data libraries' energy domain. This method faces limitations, as the resolution of calculations is restricted to volumetric cells. Analysts have to manually divide these volumes to improve detail, which can be complex and time-consuming. Additionally, Monte Carlo (MC) transport introduces statistical uncertainty, particularly in high-attenuation areas. Variance reduction (VR) methods such as Consistent Adjoint Driven Importance Sampling (CADIS) and its variants are often used to help reduce the inherent variance in MC results. A CADIS based VR technique exists that reduces the variance of the SDR by optimizing the primary step of the R2S calculation but has only been implemented for energy regimes typical of fusion devices. This work aims to improve the R2S workflow in two main ways. First, it introduces meshing capability into the special tally used in the R2S workflow, which enhances the resolution of high-energy production and destruction rate data. A complete R2S workflow is then developed, verified, and demonstrated. Second, it implements a multi-step CADIS (MS-CADIS) method for VR optimization in accelerator-driven system named high-energy GT-CADIS. The thesis further demonstrates the use of the high-energy GT-CADIS method to generate VR parameters for accelerator-driven systems considering energy regimes where cross-section libraries are not available. The high-energy GT-CADIS workflow is verified and demonstrated using test and full production models, and results indicate a significant computational speedup of up to 107 times faster convergence when compared to results obtained without VR.