The purpose of this work is to design an exact fuzzy observer for a bioprocess switching between two different metabolic states. A fed-batch baker's yeast culture is modeled by two sub-models: a respiro-fermentative state with ethanol production and a respirative state with ethanol consumption. An exact fuzzy observer model using sector nonlinearity was built for both nonlinear models; the observer gains were designed using Linear Matrix Inequalities (LMI's). The observer dynamics shows a very good tracking behavior with respect of the states of the switching partial models. The observer premise variables depend on the state variables estimated by the fuzzy observer.