Automated Semantic Layer Induction via Hierarchical LLM and RAG

A two-stage framework automates the construction of business semantic layers from raw telemetry. It uses LLM inference for high-level feature identification followed by a structured pipeline of semantic clustering and canonical naming to map noisy logs to business KPIs.