[en] In these lecture notes we explain how to build an observer for a biological system. We review the existing linear and nonlinear observers and we propose criteria to define which is the best observer with respect to the available information. Depending on the model reliability and on the level of noise, we can develop observers which use the full model description (high gain observers) or asymptotic observers which use only a mass balance model where the biological kinetics are considered as unknown inputs. If the bounds on the uncertainties can be characterised, interval observers can be designed. Each observer is illustrated with an example performed on a biological system. (author)