High-fidelity multiphysics simulations are essential for predictive digital twins in laser-based additive manufacturing (AM), yet their computational expense often restricts practical application. This work presents advances in meso- and microscale modeling of laser–material interactions, addressing two major computational bottlenecks: laser ray tracing in the melt pool and radiation absorption in laser-induced plasma plumes. First, we introduce a memory-efficient squeeze U-net surrogate, trained on fully resolved ALE3D simulations of directed energy deposition (DED) and laser powder bed fusion (LPBF) in Ti-6Al-4V [1]. The network uses local volume fraction and temperature fields to predict three-dimensional laser energy deposition on complex, evolving surfaces, including convex DED melt pools with powder shadowing and deep LPBF keyholes with multiple reflections. An energy-conserving loss function enforces physically admissible deposition and global absorptivity. This surrogate achieves approximately 97% energy conservation accuracy and accelerates the ray-tracing step by 4–40×, with inference times independent of geometric complexity and readily portable to GPUs. Second, we present a complementary study of nanosecond laser-induced plasma plumes during laser metal ablation, where a hybrid kinetic framework couples a lumped-particle direct simulation Monte Carlo (DSMC) method with either a collisional-radiative non-equilibrium plasma model or a Saha–Boltzmann equilibrium model [2]. Simulations of copper ablation in argon reveal that equilibrium assumptions can dramatically underpredict ionization, radiation absorption, and plasma shielding, while the non-equilibrium model, which resolves finite-rate excitation and ionization dynamics, shows much stronger absorption in the metal plume and agrees significantly better with experimental measurements. These results emphasize the importance of accurate energy absorption models and demonstrate how physics-informed surrogates and non-equilibrium kinetics can enable larger-scale, more predictive AM simulations. Work performed under auspices of U.S. Department of Energy by Lawrence Livermore National Laboratory contract DE-AC52- 07NA27344. License CC BY-NC-ND4.0. LLNL-ABS-2015121. REFERENCES [1] M.A. Stokes, A.N. Volkov, Z. Lin, and S.A. Khairallah, Effects of non-equilibrium ionization and excitation on radiation absorption in plasma plumes induced by ablation of metal targets with nanosecond laser pulses, J.