A natural language interface lets you describe a data-engineering problem in ordinary language, include sample data, and receive a tailored code solution. It supports conversational refinement: add requirements, request changes, or compare approaches without rewriting the implementation manually. section 2.2.1 section 1.1
A practical workflow uses the interface as an assistant within a controlled implementation loop:
- Describe the input structure, desired output columns, transformations, and business rules precisely. section 2.2.1
- Supply representative data, then inspect the generated code and output rather than treating the response as automatically correct. section 2.2.1
- Refine the conversation for additional rules such as discounts, currency conversion, or multi-step calculations. section 2.2.1
Prompt-driven order transformation
A concrete prompt asks ChatGPT to flatten nested orders into a pandas DataFrame, extract product name, brand, and color from descriptions with regular expressions, apply 20% tax to completed orders and 10% to pending orders, and skip canceled orders. The resulting table contains order, product, pricing, and quantity fields. section 2.2.1









