Traditional collaborative filtering-based recommender systems rely on prior user activity to suggest items, while deep learning-based systems focus on improving accuracy, often neglecting scalability. Moreover, these techniques are not suitable for resource-constrained environments, such as edge devices and poses significant challenges due to high computational and power requirements. This study presents an innovative approach for efficiently pruning neural networks by integrating the Lottery Ticket Hypothesis (LTH) with the Knowledge Distillation (KD) framework. These models aim to address scalability challenges in recommender systems by optimizing throughput time, query response time, power consumption while preserving accuracy. Through empirical evaluation against three baselines using Amazon dataset, the effectiveness and efficiency of these approaches were demonstrated, showcasing superior performance by having 33.21% and 43.58% reduction in MSE and MAE respectively, as compared to three baselines. Furthermore, our model consumes 25.76% less power with a 65% reduced model size compared to all the baselines.