This research manuscript presents a comprehensive methodology for optimizing supply chain performance at Walmart Fulfillment Management Services (FMS) through deep learning-based decision-making. The study begins with the acquisition of critical supply chain data, encompassing inventory levels, order histories, lead times, and external market trends, which are aggregated from diverse sources to form a robust dataset. Following data acquisition, pre-processing steps, including data normalization, are applied to ensure consistency and comparability across varying scales, thereby enhancing data quality for subsequent analyses. The Parrot Optimization Algorithm (POA) is then utilized to identify and extract optimal features that significantly influence supply chain performance, effectively reducing dimensionality and enhancing the efficiency of the predictive model. Demand forecasting is performed using the Simple Recurrent Unit (SRU) model, a deep learning technique adept at recognizing complex patterns and dependencies within time series data. This approach empowers Walmart FMS to make data-driven decisions that improve supply chain responsiveness and overall operational effectiveness, ultimately leading to increased customer satisfaction. The findings demonstrate the potential of deep learning methodologies in transforming supply chain management practices.