In modern data center networks (DCNs) high reliability, low downtime and effective management of the operations carried out is of significance. The conventional methods of maintenance and fault detection however tend to be reactive in most occasions and bring about the subject of unplanned down time and high operating costs. This paper introduces an intelligent operation and maintenance framework by using AI to overcome those challenges with fault prediction and active system monitoring based on Extreme Gradient Boosting (XGBoost). XGBoost is an efficient ensemble learning model in the gradient-boosted decision trees family, new to be resilient, scalable, and produce high predictive performance in structured data settings. The model in the proposed system is being trained on large amounts of data produced by network devices including logs, performance data, the number of errors, and the environmental conditions. Once trained, the XGBoost model effectively classifies system states and accurately predicts potential faults, enabling proactive intervention before failures occur and allowing for the automatic generation of alert messages. The application is designed to adapt over time by periodically retraining with new data, ensuring continued performance under evolving operational conditions.