Aerodynamic optimization presents several challenges due to the computational cost of CFD evaluations, and also due to the large number of design parameters. In recent times, aerodynamic optimization using machine learning (ML) techniques is being explored. However, achieving high accuracy while keeping computations efficient and cost-effective remains challenging. This paper addresses these challenges by employing a multi-fidelity data fusion method, utilizing an artificial neural network (ANN) model. High-fidelity (HF) simulations are known for their accuracy; however, they are time-consuming. On the other hand, low-fidelity (LF) simulations take lesser computational time, but are also less accurate. Therefore, the present approach combines a large number of LF data with a relatively smaller number of HF data, considering the advantages of both to perform aerodynamic optimization efficiently. The ANN model is trained to account for the complex relationships between the HF and LF datasets to predict high-fidelity outcomes with less computational effort. Further, the genetic algorithm (GA) is used as the optimizer. This methodology is applied for minimizing the drag coefficient of the Eppler E214 airfoil. The results obtained have shown an improvement in the aerodynamic characteristics by reducing the drag coefficient by 10 drag counts.