Netflix has introduced Upper, a unified domain modeling framework built into its Unified Data Architecture, to standardize domain definitions across its expanding content engineering systems. Previously, teams maintained separate domain models, which meant building custom translations between every pair of systems.This introduction of Upper provides a consistent conceptual layer, supporting engineering teams as the company grows into live events, games, advertising, and global content production.
Upper is based on W3C standards such as RDF for conceptual graph representation and SHACL for validation, and it enables the principle of “model once, represent everywhere” across the data ecosystem.Upper organizes concepts through keyed entities, their attributes, and their relationships across domain boundaries. The modeling grammar and validation structure are designed to maintain consistency as definitions evolve. Keyed concepts can be extended monotonically, allowing new attributes or relationships without modifying existing definitions allowing domains to expand over time without breaking existing models.
Graph Representation of OnePiece Domain Model (Source : Netflix Tech Blog)
As mentioned by Alexandre Bertails, principal engineer at Netflix:
Upper is designed to bootstrap itself through four foundational properties: it’s self-describing (defines what a domain model is), self-referencing (models itself as a domain), self-governing (validates itself against its own rules), and federated (closed for modification, open for extension). This self-governing foundation is what enables the governance chain that makes UDA scale.
UDA adopts a named-graph-first information model, where each named graph conforms to a governing model within the knowledge graph. The knowledge graph comprises three components: Domain Models, Data Container Representations, and Mappings, which define concepts, locate data containers, and link concepts to physical data sources. This structure provides modularity, resolution, and governance across the graph.
Mapping Between Domain Model and Data Mesh Source (Source : Netflix Tech Blog)
At the core of UDA, domain entities defined in Upper are transformed into data container representations through a process called projection. Projections generate concrete technical artifacts, including GraphQL schemas, Avro records, Apache Iceberg tables, SQL schemas, and Java types, which downstream systems can consume. UDA maintains explicit mappings between conceptual definitions and these physical data containers to ensure consistency as models evolve. Netflix engineer describes this approach as allowing changes to domain concepts to propagate automatically to all representations without manual updates, supporting both engineering workflows and operational reporting.
Domain Model to Data Containers Connection (Source : Netflix Tech Blog)
Primary Data Management (PDM) and Operational Reporting (Sphere) were among the first teams to adopt Upper. The Content Production team used UDA to track asset ingestion, approvals, and content readiness. The Advertising and Personalization team leveraged the knowledge graph to align definitions of campaigns, promotions, and targeting metadata, reducing manual reconciliation and enabling automated workflow generation. PDM exposes a taxonomy-driven interface that allows business users to navigate domain concepts without knowledge of ontology languages. Sphere generates queries automatically from the graph, minimizing the need for manual joins.
Looking ahead, Netflix plans to expand Upper and UDA to support additional projections such as Protobuf/gRPC, materialize knowledge graphs of instance data for querying and profiling, and address challenges initially posed by Graph Search. Netflix engineers aim to enhance discoverability, automation, and interoperability across Netflix’s growing content and data ecosystem.
