Tagg Fashion AI

Tagg needed to provide users with highly accurate and personalized product recommendations across a vast array of online retailers. However, the challenge lay in normalizing product data from hundreds of different sites. This included pulling essential details like pricing, stock levels, images, and accurate categorization—all of which varied significantly from site to site.

Initially, Tagg relied on Open Graph meta tags and specific HTML properties to scrape product details from various sites. However, this approach led to frequent issues, such as miscategorized data and inaccuracies whenever sites underwent redesigns or structural changes. The quality of the data was insufficient to power a robust recommendation engine, necessitating a more sophisticated solution.

To overcome these challenges, Red Foundry implemented a combination of AWS Personalize, OpenAI, and advanced scraping technology. AWS Personalize was employed to ingest user shopping behavior and generate highly personalized recommendations. Meanwhile, OpenAI was integrated to process unstructured product pages into a well-defined product model and category taxonomy, addressing the issues caused by traditional scraping methods.

By fine-tuning OpenAI’s prompt engineering, we achieved the perfect balance between token count and accuracy, ensuring that product data was accurately categorized and normalized. The advanced scraping technology continuously updated and synchronized the product database, keeping fast-moving data like stock levels and pricing accurate and up-to-date. This comprehensive solution empowered Tagg to deliver curated recommendations that captured the unique style and fashion tastes of each user.

‍