SQL Data Extraction & Cleansing

I’m working with a Microsoft SQL Server database that holds our raw operational records, and I need those tables pulled out, cleaned for data quality, and then re-loaded in a form that is immediately useful for downstream machine-learning and analytical work.

Your mission is to build a repeatable ETL workflow that:

• Extracts the source data directly from SQL Server with efficient, well-structured queries.
• Applies thorough cleaning: deduplication, missing-value handling, type standardisation, outlier checks, and consistent naming so that the dataset meets common ML requirements.
• Loads the refined data back into a staging database or delivers it as flat files (CSV, Parquet, etc.) ready to drop into my analytics notebooks.

I’m comfortable working in Python, so solutions that leverage Python, Pandas, SQL Alchemy, or even an SSIS package are all viable as long as you supply every script, configuration file, and a concise runbook explaining how to rerun the process end-to-end.

Acceptance criteria
1. No duplicate primary keys or orphaned references remain in the cleaned tables.
2. Null rates are reduced to an agreed tolerance and documented.
3. The entire pipeline runs unattended from extraction through load.

When you reply, briefly outline the toolchain you intend to use and any comparable SQL Server data-quality projects you’ve delivered.