In this presentation, we will present our recent progress in the physics-based and data-driven integrated computational materials and engineering (ICME) for metal additive manufacturing. Firstly, a lattice Boltzmann method (LBM)--based model for the ultrasound atomization process of powder generation is presented. The LBM model, which is accelerated by GPU-based computing, can correlate the powder diameter probability distribution with ultrasound parameters, such as vibration frequency and vibration magnitude. Then, we will present a multi physics processes model using a mixed sharp-diffusive interface approach. I will demonstrate how the developed model elucidates the fundamental metal AM physics (e.g., thermal history, energy absorption rate, melt pool dynamics, keyhole instability, and powder spattering) and predicts critical part quality-related quantities (e.g., defect and surface roughness). The proposed framework’s accuracy is assessed by thoroughly comparing the simulated results against experimental measurements from the National Institute of Standards and Technology AM benchmark tests and the Argonne National Laboratory using in-situ high-speed, high-energy x-ray imaging. I will also report other important quantities experiments cannot measure to show the framework's predictive capability. Finally, we integrate the multiphysics processing model with a machine learning-accelerated computer vision model to optimize cross gas flow parameters for powder spatter mitigation in laser powder bed fusion processes.