This article presents a methodology for autonomously tuning artificial neural networks using evolutionary computing, specifically genetic algorithms. The proposed approach aims to automate the calibration of hyperparameters and network architectures, which is traditionally manual and expertise-intensive. Genetic operators such as selection, crossover, and mutation are applied to evolve architectures that optimize model performance by encoding network configurations as genotypes. The methodology is experimentally validated on classification datasets, showing significant improvements over traditional manual tuning and random search methods. Results demonstrate the effectiveness of the evolutionary approach in constructing more adaptive and efficient neural networks, with potential applications in scenarios where human intervention is limited or costly.