This thesis explores the application of reinforcement learning to perishable inventory management, focusing on optimizing ordering and discounting decisions to enhance profitability and reduce waste. Traditional inventory policies often fail to adapt dynamically to demand fluctuations and product perishability, necessitating more flexible, data-driven approaches. We present a formulation of the problem as a Markov Decision Process and investigate various reinforcement learning techniques, including Least-Squares Policy Iteration and rollout-based methods. Computational experiments demonstrate that rollout-based approaches outperform traditional heuristics, yielding profitability improvements of 0.8% to 5.1% and substantial waste reduction. Notably, waste reduction is particularly effective in cases with low demand variability.