Transforming complex, unformatted archives, bibliographic records, and retail data into unified, queryable search systems using Claude reasoning.
Using Claude's high-context reasoning to align legacy records, ambiguous publishers, and unstructured descriptions into normalized relational database schemas.
Eliminating duplicates across heterogeneous catalog archives without brittle regex rules, identifying matching items even with erratic naming conventions.
Combining local SQLite indexes, FTS5 engines, and real-time LLM-enriched search filters for lightning-fast catalog querying.
Architecture & Stack
Python ETL Pipelines · SQLite FTS5 · Meilisearch · Anthropic Claude 3.5 API