Airfoil shape optimization is key to improving aerodynamic performance in aerospace applications. This MSc thesis investigates two machine learning algorithms, Factorization Machine Quantum Annealing (FMQA) and single-step Deep Reinforcement Learning (sDRL) in the context of airfoil shape optimization, and compares them with a standard SciPy optimizer. The algorithms are coupled with a Lattice Boltzmann Method solver and tested across multiple aerodynamic objectives, flow regimes, and parameterizations, including classical schemes and a novel Variational Autoencoder (VAE)–based representation. Results show that FMQA and sDRL consistently identified high-performing airfoils. sDRL often achieved slightly better aerodynamic objectives, while FMQA outperformed in specific cases. Compared with SciPy, the machine learning methods produced competitive or superior results depending on the scenario. The VAE scheme improved convergence and performance across all methods. These findings highlight FMQA and sDRL as flexible and competitive strategies for aerodynamic design problems.