The fed-batch mode of bioreactor operation is commonly used in the chemical and biopharmaceutical industries. Bioreactors typically are characterized by highly nonlinear behavior occurring on both a macroscopic reactor scale and a microscopic cellular scale. This nonlinearity, coupled with the dynamic nature of the fed-batch mode, makes for a challenging control problem. In this work, a multiscale model(1) describing the growth of yeast in an aerobic fed-batch mode is employed to address both the offline and online optimization and control issues for the fed-batch fermentation problem. Nonlinear optimization techniques are used offline to generate an optimal substrate feed profile for the nominal problem that maximizes the end of the batch ethanol concentration; shrinking-horizon nonlinear quadratic dynamic matrix control is used online for closed-loop trajectory tracking and disturbance rejection. Nominal performance is admirable, resulting in final ethanol concentrations within 1% of the open-loop optimal ethanol concentration. An online re-optimization strategy is presented for disturbance compensation, which improves the final ethanol concentration in the presence of disturbances by 17% when compared to the case when no disturbance compensation strategy is employed.