Worked on the Unstructured-IO/unstructured-ingest repository to enhance the reliability of SQL source connectors by addressing a recurring KeyError during data ingestion. Applied Python and SQL expertise to ensure the id_column is always included in SELECT statements, preventing runtime errors across Snowflake, Teradata, Postgres, SQLite, and SingleStore connectors. Introduced case-insensitive column resolution in the base SQLDownloader, improving compatibility with databases that return column names in varying cases. Expanded and validated unit test coverage, with all existing and new tests passing, which reinforced CI reliability. This work contributed to more stable data processing and reduced downtime in production environments.