This study presents the optimal design of an 18-passenger air taxi through the integration of classical aircraft design methods and a Q-learning based reinforcement learning framework. Initially, baseline parameters, including maximum takeoff weight, empty weight, fuel consumption, and wing area, were estimated using established classical conceptual design relations. A simulation environment was then developed in which the state space was defined by key nondimensional ratios
while the action space consisted of bounded adjustments to wing area, aspect ratio, and thrust/weight ratio. The agent was trained for 2000 episodes with a maximum of 40 design steps per episode using an ϵ\epsilonϵ-greedy policy under a fixed mission scenario defined by prescribed cruise altitude, cruise speed, range requirement, and field length constraints. The optimization process resulted in a reduction in wing area (3.8% decrease) and an increase in aspect ratio (2.10%). The horizontal and vertical tail areas were reduced by 2.9% and 3.2%, respectively. These geometric refinements led to reductions in empty weight (4.1% decrease) and fuel consumption (5% decrease). Aerodynamic and performance improvements were achieved, including increases in lift/drag ratio (20% increase), rate of climb (33.3% increase), and cruise speed (6.7% increase). Consequently, the flight range improved (14.5% increase). The findings demonstrate that integrating classical design methodologies with reinforcement learning provides an effective data-driven framework for simultaneous weight reduction, aerodynamic efficiency enhancement, and mission performance improvement in air taxi aircraft design.