Increased information intensity is greatly expanding the importance of model-based decision-making in a variety of areas. This paper reviews applications that involve large-scale combinatorics, data uncertainty, and game theoretic considerations and describes three related algorithm architectures that address these features. In particular, highly customized mathematical programming architectures are discussed for time-based problems involving significant combinatorial character. These architectures are then embedded into a simulation-based optimization (SIMOPT) architecture to address both combinatorial character and significant data uncertainty. Multiple SIMOPT objects are combined using a coordinating architecture to address game theoretic issues. A key focus of the paper is a discussion of the algorithm engineering principles necessary to mitigate the NP-complete nature of practical problems. The life cycle issues of algorithm delivery, control, support, and extensibility are important to sustained use of advanced decision-making technology. An inductive development methodology provides a means for developing sophisticated algorithms that become increasingly powerful as they are subjected to new constraint combinations. Implicit generation of formulations is crucial to routine large-scale use of mathematical programming based architectures.