Zeta Global has developed a sophisticated data architecture to support its AI-driven marketing initiatives. Following the acquisitions of LiveIntent and Marigold, Zeta faced the challenge of managing a diverse array of data sources and processing patterns. The company moved away from traditional data warehousing to a more flexible architecture that can efficiently handle various data types and workloads. This composable design enables separate components to evolve independently, allowing Zeta to adapt quickly without overhauling the entire system. The core of Zeta's data platform includes object storage with Parquet format, lakehouse table formats for transactional guarantees, and a metadata catalog that governs data discoverability and governance. They chose Iceberg as the table format and implemented S3Table with AWS Glue Catalog for managing Iceberg tables. This combination improved operational efficiency and allowed for scalable multi-tenant ingestion. To further enhance data access, Zeta employs a multi-account federated catalog architecture, which maintains autonomy at each account level while providing secure access to shared datasets without duplicating data. Looking ahead, Zeta aims to strengthen its architecture by establishing a unified data product framework to standardize dataset management and improve interoperability. They are also focused on scaling infrastructure significantly to handle increased customer adoption, targeting enhanced performance across storage, catalog services, and compute orchestration. This strategic direction supports their mission of delivering advanced marketing solutions that leverage customer intelligence effectively.