Mead is an alcoholic beverage fermented from water and honey produced by the action of yeasts, typically strains of Saccharomyces cerevisiae, on carbohydrates such as glucose and fructose. Producers often face challenges stemming from limited knowledge and lack of control over important process parameters. The present work aims to develop a virtual sensor based on Artificial Neural Networks capable of predicting the concentration of cells (X), total sugars (S) and ethanol (P). during the fermentation to produce mead with the addition of Jabuticaba peel pulp. To make this possible, the chosen input variables will be operational temperature (T), culture media pH, concentration of total soluble solids (°Brix), and optical density (D.O.). For the acquisition and simulation of the feedforward network with supervised training, experimental data from a fermentation conducted during Costa's scientific initiation project (2020) (FAPESP Process: 19/24444-1) will be used. The neural networks will be evaluated in various configurations, and the identification of the best training algorithms, activation functions, number of intermediate layers, and the number of neurons in each layer will be made to optimize the prediction of the output variables by the network.