Unit price optimization is a critical element of competitiveness and profitability maximization in the retail industry. The present research paper presents a machine learning approach to predict the optimal unit price for goods considering various impacting variables. Past sales data are collected and preprocessed, with feature engineering techniques for enhancing the dataset. Freight price, total customers, seasonality, and competitor prices are some of the key features employed in the model product ratings, and quantity of sales. A neural network model is developed and trained using TensorFlow and Keras frameworks. The performance of the model is assessed using the Mean Absolute Error (MAE) metric, which is 0.136. This low error rate indicates a high accuracy in forecasting the best prices. In order to make this model accessible to retail managers and facilitate real-time pricing decisions, the model is integrated in a Streamlit application. The web-based user interface allows users to input data that is relevant and receive immediate price forecasts, thus supporting dynamic and data-driven price adjustments. The application is made available to facilitate easy use by non-technical users so that the tool can be used to improve their pricing behavior. This entire methodology, taking several important predictors in one model, has the capability to offer a strong solution for retail price optimization. Using sophisticated machine learning algorithms and a simple-to-deploy platform, this project seeks to enhance the process of retail price decision making, which ultimately will result in better sales performance and profitability.