Inspiration -- help simplify the shopping experience to alleviate choice and information overload for busy consumers searching for high-consideration items
What it does -- uses LLM to generate question and answer options that effectively capture consumer's preferences, then feed those preferences back into the LLM to curate personalized product recommendations
How we built it -- 1) Scraped Amazon for products (in JSON format) and their associated reviews (in CSV) and uploaded them into an S3 bucket. 2) Used S3 bucket to create a knowledge base in Amazon Bedrock. 3) Used that knowledge base with a Claude v2 foundational model to handle engineered prompts.
Challenges we ran into: 1) Amazon Bedrock failed to respond to incompatibilities with sufficient explanation -- just said "can't handle that". 2) We couldn't scrape products due to Amazon bots limiting our ability to do so. 3) The Workshop Studio had limited permissions.
Accomplishments that we're proud of -- building something that works
What we learned -- Bedrock knowledge base integration
What's next for PersonalAIzed Product Discovery -- scaling to additional products beyond the two we started with (baby strollers and coffee machines)
Log in or sign up for Devpost to join the conversation.