This talk will discuss the application of linear observer theory to engine control with a specific focus on observers based on exhaust measurements. A key aspect is the model of the process between the fuel injector and the exhaust measurement of the fuel-air ratio. The motivation for this architecture is that a direct measurement of the relevant quantity (the exhaust chemistry entering the catalyst) will typically lead to the most accurate control. An interesting aspect of the control design is that linear theory can be applied to a very nonlinear problem. In fact, the sensor in one case is on-off and yet the linear observer does an excellent job of reconstructing the state. The key idea to keep in mind for a such a design is to match the process dynamics with the observer model, whether it be linear or not. The design process and the experimental evaluations of the performance of the scheme on various engines is described.