Developed a deep learning-based sales forecasting system to predict future product demand and support data-driven inventory planning. The project focused on learning complex sales patterns from historical retail data, enabling more accurate demand forecasting and helping businesses optimize stock management and operational planning. I designed an end-to-end forecasting pipeline that included data preprocessing, feature engineering, model development, evaluation, and prediction. A deep neural network incorporating product and store embeddings was trained to capture relationships between categorical and numerical features, allowing the model to learn purchasing patterns across different stores and products. The solution was evaluated against traditional machine learning models to assess forecasting performance. Project Highlights: Developed a deep learning forecasting model using product and store embeddings to improve demand prediction accuracy. Trained and evaluated the model on a retail dataset containing 150,150 sales records. Achieved an R² score of 0.9771 with a Mean Absolute Error (MAE) of 1.96, demonstrating high predictive performance. Compared the deep learning model with traditional machine learning approaches, including Random Forest and XGBoost, to validate performance improvements. Built a complete forecasting pipeline covering data preprocessing, feature engineering, model training, evaluation, and prediction.