StarFlow
The Tiangong™ pharmaceutical matching engine continues to advance, from “Manual · Zero Start” to AI-driven, highly accurate matching. Its six-generation evolution spans morphological matching, dynamic scoring, pharmaceutical small models, LLM Embedding, and RAG Rerank.
The six-generation evolution of Tiangong™.
Across six generations, each generation continues training on the cleansing results of its predecessor, steadily improving accuracy.
Note: The chart above illustrates the relative evolution of capabilities across generations of the Tiangong™ engine to explain the technology's development. It does not represent specific accuracy figures.
One engine. Four major advances.
- Real-time cleansing and alignment
- Millisecond-level anomaly detection
- No T+1 batch runs required
- Collection → governance → ready for use in one stream
- LLM Embedding semantic matching
- RAG + Rerank optimization
- 5-level result classification
- Pharmaceutical-specific small models
- Knowledge graph semantic alignment
- Unification of heterogeneous sources
- Structured + unstructured data
- SKU × organization × time
- Significantly reduce manual operations and maintenance
- Data Cleansing Automation
- Automatic anomaly remediation
- No on-site staffing required
Note: The product capabilities and outcomes described above are general introductions to functionality. Actual results vary depending on data conditions, deployment environments, and business scenarios.