Huuray, a global provider of digital rewards and gift cards, sought to enhance their GifTrix platform—a solution designed to offer personalized corporate gifts to employees. The challenge was to efficiently select 20 relevant products from an extensive catalog of over 50,000 items, ensuring each gift resonated with individual employee preferences.

Traditional methods of gift selection were inadequate for handling the vast product range and the need for personalization. Key challenges included:

• Determining product relevance for diverse personas (e.g., toddlers, mid-age DIY enthusiasts).
  For example, humans inherently understand that a garden swing is meant for children aged 4-12, even though this information might not be explicitly mentioned in the product description.

• Inability to process 50,000+ products simultaneously through AI for relevance assessment. The token context window simply would not make that possible - or feasible from a cost perspective.

• Lack of metadata on products, such as age suitability or interest alignment.