Expedia Group is enhancing its Machine Learning capabilities with a centralized vector embedding service, addressing the growing demand for vector similarity search due to advancements in AI. Vector embeddings allow for comparing disparate inputs by representing data as numerical vectors. The service aims to streamline the process for developers by reducing the complexity of integrating and managing these embeddings, which improves collaboration across teams. The Embedding Store Service integrates with Feast, an open-source feature store, to facilitate the management and querying of vector embeddings at scale. It offers essential functionalities like CRUD operations for embeddings, similarity search capabilities, and filtering by metadata, including model and version. This setup ensures data consistency and eases the discoverability of embeddings, allowing users to efficiently locate and utilize existing data tailored to specific needs. Feast also supports an online-offline store model. The online store is optimized for real-time queries, making it suitable for applications like recommendation systems. The offline store archives historical data, enabling batch processing and model training. This dual system allows for a seamless flow of data, ensuring that both current and historical embeddings are readily accessible, fulfilling various analytical needs while maintaining traceability.