"We replaced our legacy ETL with DuckDB for in-process analytics on large Parquet datasets. We achieved clinical insights by querying data directly from data lake without expensive data movement. This shift enabled real-time patient risk scoring while reducing infrastructure costs."