Aglar watershed, located in the Tehri-Garhwal district of Uttarakhand, is a part of Lesser Himalaya. The diverse geography, highly rugged topography, and continuously changing climatic conditions make the Himalayan region landslide-prone. In the Aglar watershed, landslides occur frequently due to various geological, anthropogenic, and vegetation factors. Reducing landslide risk requires an accurate assessment of landslide susceptibility through mapping. This paper delves into Extreme Gradient Boosting (XGBoost), a well-known machine-learning algorithm for landslide Predictive Modeling. In addition, XGBoost is an ensemble learning algorithm that combines numerous weak predictive models (decision trees) to develop a strong landslide predictive model. However, the effectiveness of algorithms greatly relies on choosing the best Hyperparameter values. This study examines the utilization of hyper-parameter optimization techniques in combination with the XGBoost algorithm to improve the performance of landslide susceptibility Modeling. Sixteen causative factors were utilized in this study based on the topographical and geological conditions of the study area. Including active, old, and stabilized landslides, 375 inventories were created using satellite imagery and validated through field investigation. The inventory map was divided into training (70%) and testing (30%) datasets through a random sampling method and further converted the dataset into dichotomous binary data to construct the XGBoost Landslide prediction model. Higher classification accuracy with an AUC-ROC curve (AUC = 0.92) could be obtained from the XGBoost algorithm. The result shows that with the tuning of Hyper-parameter optimization, the XGBoost model gains the 84.73% kappa score- and 92.40% accuracy score, which generates the final Landslide Susceptibility Mapping of the study area.