This work presents a computationally efficient framework for high-fidelity part-scale thermo-mechanical simulations in Metal Additive Manufacturing (MAM), specifically focusing on Direct Energy Deposition (DED). MAM enables precise layer-by-layer fabrication of complex metal components, offering significant design flexibility and material efficiency. However, challenges such as intricate thermal gradients, evolving residual stresses, and complex material behavior can impact the structural integrity and performance of the final component. Accurate predictive modeling is essential for optimizing process parameters and ensuring part reliability. Standard Full-Order Finite Element (FE) simulations, while capable of capturing detailed physics, are often computationally expensive and impractical for large-scale or iterative process optimizations. To address these challenges, this work introduces a novel local-global simulation strategy leveraging the Virtual Domain Approximation (VDA) method [1]. The proposed approach strategically reduces computational costs by restricting the numerical domain to the Heat-Affected Zone (HAZ), where critical thermal and mechanical interactions occur. The residual-based VDA technique ensures accurate approximation of boundary conditions, effectively coupling the local and global solution spaces while preserving the accuracy of mechanical predictions. A comparative evaluation against standard FE simulations demonstrates that the proposed method achieves significant computational speed-up without compromising accuracy. By enabling efficient, high-resolution analysis of MAM-induced thermal and mechanical effects, this work provides a powerful tool for process optimization and part qualification. The developed framework supports the broader industrial adoption of MAM technologies, particularly in high-performance sectors such as aerospace, automotive, and energy, where reliability, precision, and efficiency are paramount. [1] Moreira2025_GLApproach} Moreira, Carlos A. and Chiumenti, Michele and Caicedo, Manuel A. and Baiges, Joan and Cervera, Miguel, High-Fidelity Part-Scale Simulations in Metal Additive Manufacturing Using a Computationally Efficient and Accurate Approach. Under review https://ssrn.com/abstract=5020066 or http://dx.doi.org/10.2139/ssrn.5020066