The challenge
Keyword search was not enough for business users who described data needs in natural language and often did not know the exact dataset, table or terminology to search for.
RAG / Search
RAG Semantic Search for Enterprise Data Discovery
A retrieval-augmented semantic search experience designed to help enterprise users find relevant governed data assets faster.
Role
AI Engineer
Context
Global Supply Chain Enterprise
Period
2023–2026
What I built
Implemented semantic retrieval over enterprise metadata using embeddings, metadata filtering and relevance-focused ranking. The experience translated natural-language intent into better matches while preserving governance context and source traceability.
Architecture
- Semantic chunking of metadata-rich records
- Embedding-based retrieval
- Metadata filters for business domain and ownership
- Reranking and relevance evaluation
- Source-aware responses and citations
Outcomes
- Approximately 50% reduction in data-discovery time
- More relevant results for natural-language queries
- Improved discoverability of governed enterprise data assets
Technology
PythonEmbeddingsVector SearchRAGSQLEnterprise Metadata