iQIYI Data Middle Platform: Architecture, Capabilities, and Future Outlook

With the market’s increasing recognition of data value, the explosive growth of data across industries and the rapid evolution of big‑data infrastructure have introduced many new data problems. iQIYI builds a data middle platform to centralize, standardize, and service‑orient the data capabilities, enabling business development and deeper AI‑driven value extraction.

Definition and Positioning : The data middle platform abstracts common data capabilities, providing unified governance and asset management, lowering the barrier for business units to leverage data for value creation.

Key Challenges Addressed include fast‑changing infrastructure, massive data volume, integration of AI, high‑frequency data freshness, inconsistent data definitions, quality reliability, high usage cost, resource waste, chaotic asset management, high integration cost, poor data flow, and long production pipelines.

Eight‑point Solution :

One‑stop data development platform.

Unified metadata center.

Configurable data service hosting.

Full‑life‑cycle data governance.

Standardized data production and collection.

Data lake and warehouse ecosystem.

Centralized data exchange.

AI‑driven automation for repetitive tasks.

Platform Architecture consists of a bottom‑layer big‑data infrastructure (cloud‑enabled), a unified data production and ingestion layer, a unified development platform, a data lake/warehouse layer, and a unified service layer that is elastic, configurable, monitorable, and composable.

Unified Service Layer tackles reuse of data and APIs, data discovery, standardization, efficient data ingestion, and service usage audit.

Data Service Capabilities include one‑stop development and operation, full‑link traceability, API marketplace, pull/push access modes, logical data models, and a data gateway with authentication, rate‑limiting, and monitoring.

Metadata Service provides search and graph‑based exploration of data assets, including definitions, owners, quality metrics, lineage, and popularity, powered by a metadata center that aggregates technical and business metadata.

Inference Service integrates AI models trained offline into the online prediction platform, offering real‑time inference via REST or RPC interfaces with features such as pre‑loading, hot‑updates, and A/B testing.

Future Outlook focuses on large‑scale cross‑source data merging, composable service interfaces, model‑driven service creation, and automated, template‑based service definition to reduce manual effort.

The presentation concludes with a summary of iQIYI’s data middle platform achievements and a call for audience engagement.