Accurate State of Health (SOH) and Remaining Useful Life (RUL) estimation of lithium-ion batteries is important to guarantee their longevity, reliability, and safe operation in electric vehicles, mobile phones, and energy storage systems. Traditional estimation techniques like statistical modeling and deep learning are known to require vast computational resources and data. In this work, we introduce a complete machine learning-based method using the Improvised eXtreme Gradient Boosting (I-XGBoost) algorithm, optimized using Bayesian optimization, for precise yet computationally effective SOH and RUL estimation. For better model interpretability, we use explainable AI (XAI) tools such as Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive Explanations (SHAP), which give better insights into feature importance and decision-making. The model is trained and validated on NASA Ames Prognostics Data Repository battery cycling data, specifically B0005, B0006, and B0018 datasets. Experimental results demonstrate that the designed approach exhibits high predictive accuracy, computational efficiency, and robust interpretability, surpassing traditional approaches. This paper demonstrates the potential of XGBoost, coupled with XAI techniques, as a realistic and scalable approach for real-time battery health monitoring and predictive maintenance applications.