Crystallization plays a vital role in separation and purification methods for the production of therapeutic drugs. Considering the fact that crystal size and shape distributions have a significant influence on the bioavailability of drugs such as the dissolution rate, filterability, and stability as a carrier to the target site, the production of crystals with desired size and shape distributions is of particular interest to the pharmaceutical industry. Motivated by these considerations, this dissertation focuses on the development of a multiscale modeling and simulation framework for crystallization processes that elucidates the relationship between molecular-level processes like crystal nucleation, growth and aggregation and macroscopically-observable process behavior and allows computing optimal design and operation conditions. Using protein crystallization as a model system, the multiscale framework encompasses: a) equilibrium Monte-Carlo modeling for computing solid-liquid phase diagrams and determining initial crystallization conditions that favor crystal nucleation, b) kinetic Monte-Carlo modeling for simulating crystal growth and aggregation and predicting the evolution of crystal shape distribution, and c) integrated multiscale computation linking molecular-level models and continuous-phase macroscopic equations, covering both batch and continuous crystallization systems. The multiscale model parameters and predictions are calibrated and tested with respect to available experimental data. Then, this dissertation addresses model predictive controller designs that utilize the insights and results from the multiscale modeling work and real-time measurements of solute concentration and temperature to manipulate crystallizer conditions that lead to the production of crystals with desired size and shape distributions. To enhance the ability of the predictive controller to deal with batch-to-batch parametric drifts, a common problem in industrial crystallization owing to changes, for example, in the pH level or impurity concentration in the feedstock container, a run-to-run-based model parameter estimation scheme will be presented that uses moving horizon estimation principles to update the predictive controller model parameters after each batch and leads to the consistent production of crystals of desired shape at the end of each batch.