<p>This study investigates the use of artificial intelligence (AI), in particular neural networks, to predict the electronic properties of quantum dots (QDs) with cylindrical, spherical, and toroidal geometries. A hybrid approach combining the finite difference method (FDM) to solve the Schrödinger equation and artificial intelligence models was used to predict the electronic properties of the quantum dot. The neural networks were adapted to each QD geometry and optimized using modern techniques such as data normalization, the Adam optimizer, and five-fold cross-validation to ensure robustness and generalization across limited datasets. The k-fold validation strategy provided a statistically reliable assessment of model performance by repeatedly partitioning the data into complementary training and validation subsets, thereby minimizing the risk of overfitting and enhancing predictive stability. The results demonstrated the ability of the models to provide accurate predictions under various geometrical and external conditions, including the influence of electric field and off-center shallow donor impurity. The study highlights the role of AI in resolving the limitations of traditional computational methods, offering a faster and more efficient alternative for modeling QDs. This advance provides a better understanding of quantum confinement effects and paves the way for practical applications in optoelectronics, photonics, and quantum devices.</p>