Building an Intelligent Data Analysis Platform Based on a Unified Semantic Layer

The presentation introduces the business context of Xiaomi, highlighting the wide range of product lines, complex data sources, and strict security requirements that drive the need for a unified data service and analysis platform.

The system architecture positions the data service platform at the core of the data middle‑platform, with a data development layer (storage and compute), a unified data warehouse, and two main product groups: the data service platform and the intelligent data analysis platform.

The platform’s key modules include:

Data assets: integration of various source types (online business data, MySQL, commercial databases, data warehouse, Hive/Iceberg tables) and pipelines for real‑time and batch ingestion.

Unified semantic layer: a semantic model that supports view, dimension, and metric definitions, row‑level security, automatic acceleration, and both SQL and MDX query execution.

Intelligent analysis: anomaly detection, root‑cause analysis, and predictive alerts that turn data into actionable insights.

Advanced features such as multi‑engine adaptation, SQL optimization, automatic modeling assistance, and engine selection (Doris vs. ClickHouse) further enhance performance and usability.

Product capabilities are demonstrated through rich chart types, customizable reports, proactive alerts, low‑code embedding of dashboards into business systems, and data insight visualizations that combine anomaly detection with multi‑dimensional root‑cause analysis.

The session concludes with a thank‑you note to the audience.