Laser Powder Bed Fusion (LPBF) is the most widely used metal Additive Manufacturing (AM) process because of its unique capabilities of building complex geometries without extensive post-treatment. However, an in-depth understanding of the LPBF process to manufacture the quality part is still under development as the involved processes are highly dynamic, consisting of micrometer (μm) variation within the time span of milliseconds (ms). Two key focuses in developing a comprehensive understanding of the LPBF process are the evolution of the melt pool and keyhole. Experiments to study and analyze the melt pool and keyhole dynamics are limited by time, effort, cost, and resources required. Computational analysis of the LPBF process can significantly aid in accelerating the investigation. We developed a C++ Finite Volume Method (FVM) code to solve the Navier-Stokes equation to accurately model the LPBF process for different conditions (laser power, scan speed, spot radius). Combined with the physics-based melt pool prediction, the inherent physical characteristics of input energy from the laser source can be calibrated based on machine variation and manufacturing environments. The developed code has successfully demonstrated the capability of such an approach by predicting the 2022 NIST AM Bench Challenge problems Utilizing the physical understanding of laser heating sources and converting them into energy input, a physics-guided heat source significantly accelerates the computational methods but with minimal sacrifice in accuracy and thus allows the capability of computationally modeling full-scale AM parts. It has been demonstrated that the appropriate selection of a laser heating source helps reveal important information about the dynamics of the melt pool for different LPBF process conditions. Further developing the technology, a stochastic computational framework based on Markov Chain Monte Carlo (MCMC) was also demonstrated to model the uncertainty in the LPBF process. The capability of predicting the uncertainty in the AM process eventually helps predict the surface roughness and lack of fusion porosity in part-scale AM manufacturing. Thus, the developed codebase is an effective tool for the hybrid twin framework of the LPBF AM machine. Future development of this technology focuses on efficiently reducing the computational modeling complexity with an aim to model and predict part-scale surface roughness, porosity, mechanical property predictions, and fatigue performances.