Personalized movie recommendations have become a core feature in many online streaming platforms. Conventional recommendation techniques, including collaborative filtering and content-based methods, frequently encounter challenges in handling sparse datasets and inadequately model the intricate interactions between users and movies. These systems typically rely on user ratings or metadata, which can be limited and not reflective of underlying connections within user-item interactions. As a result, they may miss key patterns, leading to less accurate recommendations, particularly for new or less active users. To address mentioned restrictions, Graph Neural Networks (GNNs) based implementation is presented for personalized movie recommendations through link prediction. This allows for the discovery of latent relationships between users and movies, significantly improving recommendation accuracy. The GNN-based model uses link prediction to identify potential new connections between users and movies, predicting the likelihood of a user enjoying a particular movie. Through extensive experimentation, we demonstrate that our graph neural approach yields more personalized and accurate recommendations and provides a promising performance of 0.9331 AUC score on the standard dataset.