The multitude of interacting variables complicates real-world decision-making. Deterministic 'classical' decision-making models are inadequate for the stochastic models present in the modern world. This paper offers a multi-level dynamic probabilistic modeling (MDPM) framework that integrates a probabilistic model witha operational research (OR) strategies and decision-making (intelligent and computational) automations. The framework combines dynamic Bayesian networks, Markov decision processes, and multi-level stochastic programming with metaheuristic and machine learning approaches (genetic algorithms, simulated annealing, reinforcement learning) to design intelligent, scalable, and flexible solutions in diverse applications. The framework's foundations, mathematical models, algorithms, and applications are presented. The MDPM case studies show that the framework is superior to the traditional single-level different approaches because it provides accurate results, is stable with respect to uncertainty, and provides superior optimization. The case studies serve as a comprehensive, rational, and computationally lightweight development for decision-making models for the supply chain, healthcare, smart grids.