Part 5/11:

• Prompt Engineering & Generating Tags: Using GPT-4, she crafted prompts to generate tentative "life stage" tags for each product, such as "family with infant," "teenager," or "senior." Subsequently, expert review ensured the accuracy and relevance of these tags.

• Model Fine-Tuning: The refined data was then used to fine-tune a smaller language model, Mistal 7B, enabling the system to predict life stage tags for the entire product catalog automatically.

This method allowed the system to classify 45,000 products with reasoned tags, significantly reducing manual effort while improving consistency.

Mapping Customer Transactions to Product Metadata

The next step was aligning transactional data with the product metadata: