Near infrared spectroscopy (NIR) uses fiber-optics for rapid data transmission, is robust, simple, and sensitive at both low and high solution concentrations. Therefore, it is particularly suitable for monitoring industrial processes. This study investigates the use of NIR for monitoring batch cooling crystallization processes, and emphasis is placed on applying genetic algorithm (GA) for wavelength selection in partial least squares calibration model development. The calibration data was collected for under-saturated and saturated solutions, as well as for alpha- and beta-form crystal slurries of L-glutamic acid at a variety of solution concentrations, temperatures, and solid concentrations and sizes. The GA method proves to be capable of effectively selecting a small number of wavelengths and the models thus developed give improved prediction performance in terms of generalization capability compared to models derived using the full spectrum. The developed models are successfully applied to monitoring batch cooling crystallization of L-glutamic acid under seeded and unseeded conditions and with varied cooling rates.