As Non-Fungible Tokens (NFTs) are booming and gaining popularity, but the traditional recommendation systems out there are falling short. These systems struggle with the lack of semantic understanding, cold start issues, and sparse data. In this research, we propose an advanced recommendation system using Sentence Transformers and Cosine Similarity. This system converts the NFT metadata into high-dimensional embeddings for making semantic-rich recommendations. The whole process involves preparing the NFT data, using the "all-MiniLM-L6-v2" model to create highdimensional embeddings, and at last ranking the NFTs using Cosine Similarity. It is designed to be scalable, hence can handle large datasets and also provides personalized discovery and fits perfectly into the NFT marketplace. This study enhances the NFT recommendation systems by using the Sentence transformer and Cosine Similarity, which is a deep-learning approach