Accurate prediction of long-term mass diffusion in metallic alloys is essential for the rational design of alloy compositions and microstructures. Bridging the enormous gap between atomic vibration timescales and experimentally observable diffusion timescales remains a formidable challenge. This work presents a novel computational framework that achieves this, operating purely from an interatomic potential without any continuum-level assumptions. The core methodology combines Gaussian Phase Packets (GPP) [1] for finite-temperature quasistatic atomic relaxation with Harmonic Transition State Theory [2] to compute local environment-dependent transition rates. Nudged Elastic Band (NEB) calculations [3] are performed on-the-fly on uniquely identified atomic environments to obtain energy barriers, vacancy segregation free energies, and solute-vacancy interaction energies. This staggered scheme alternates between concentration updates via a master equation and GPP thermal relaxation, enabling the simulation of vacancy-mediated substitutional diffusion at realistic timescales. For dilute alloys, the framework [4] is validated by reproducing the self-diffusion coefficient in copper against prior experimental data [5], and by simulating magnesium segregation to a stacking fault and a symmetric tilt grain boundary in aluminum. The results show excellent agreement with Langmuir-McLean isotherms and span timescales from seconds at 600 K to over 10^5 years at 300 K, within hours of computational time. For non-dilute alloys, a reduced-order Markovian interdiffusion model is derived [4] by eliminating the fast vacancy dynamics analytically from the concentration evolution equations, yielding a microstructure-dependent atomistic analog of Darken's equations. This is validated by simulating copper/alpha-brass interdiffusion across four crystallographic interfaces at 1058 K, showing marker displacement kinetics in strong agreement with the classic Kirkendall experiments. Collectively, this framework enables atomistic simulations of non-equilibrium mass diffusion at engineering timescales, opening new avenues for studying non-equilibrium mass transport, microstructural evolution, and long-term alloy degradation from first principles.