This paper presents a fuzzy inference system (FIS) enhanced through genetic algorithms (GAs) to predict the optimal harvest time for Stevia rebaudiana based on climatological data. The proposed Mamdani-type FIS integrates five key meteorological variables—solar radiation, humidity, dew point, total precipitation, and average precipitation—to estimate the °Brix value, which indicates the crop’s sweetness and readiness for harvest. Three types of membership functions (triangular, trapezoidal, and Gaussian) were optimized using evolutionary encoding strategies, where each chromosome encodes the type and parameters of the membership functions. A total of 90 parameters were evolved across 20 generations using selection, crossover, and mutation operators. The system’s rules were derived from expert knowledge and refined with data-driven insights. Validation was conducted using an 80-20 hold-out strategy and evaluated through metrics such as mean squared error (MSE) and accuracy. Results demonstrate that the trapezoidal configuration achieved the highest accuracy (96.43%) and the lowest MSE (0.0357), outperforming both triangular and Gaussian configurations. These findings validate the effectiveness of integrating fuzzy logic and evolutionary algorithms in precision agriculture, offering a non-destructive, interpretable, and robust method for harvest prediction.