In this paper, we develop a Model-Based Geometric Control Algorithm (MGA) for controlling the dissolved oxygen concentration in fermentation processes. The algorithm is developed on a generic system description which encompasses a wide range of models commonly used to describe bioprocesses. Consequently, the algorithm we have derived will apply to any process whose model fits the generic description. The algorithm uses information contained in the shape (geometry) of the profile of the state variables as they evolve in time to adapt to process variations. There are two components in the algorithm, an estimator and a controller, whose functions complement each other. The estimator component of the algorithm predicts the states and the parameters of the system one time step ahead. Since the estimator is deadbeat, the algorithm converges in a finite number of steps. The control component of the algorithm uses the states predicted by the estimator and executes a control action so that predicted error falls below a desired level. Simulations for comparing the performance of the IMC, PI and MGA controllers are presented. The MGA was implemented on-line to control the dissolved oxygen in an aminoglycoside antibiotic production process by a Streptomyces and the results of its performance are also presented.