Reliable control of the specific growth rate (μ) in fed-batch fermentations depends on the availability of accurate online estimations of the controlled variable. Due to difficulties in measuring biomass, μ is typically estimated using reference models relating measurements of substrate consumption or oxygen uptake rate to biomass growth. However, as culture conditions vary, these models are adapted dynamically, resulting in complex algorithms that lack the necessary robustness for industrial applicability. A simpler approach is presented where biomass is monitored using dielectric spectroscopy. The measurements are subjected to online balances and reconciled in real time against metabolite concentrations and off-gas composition. The reconciled biomass values serve to estimate the growth rate and a simple control scheme is implemented to maintain the desired value of μ. The methodology is developed with the yeast Kluyveromyces marxianus, tested for disturbance rejection and validated with two other strains. It is applicable to other cellular systems with minor modifications.