Working Backwards AI is a generative-AI platform that simulates feedback from virtual customers and domain experts to help Product Managers craft better ideas, refine product concepts, and write PR/FAQ documents faster. I led the end-to-end design strategy—from defining multi-agent interaction models to building an AI-native design system that visualizes reasoning and feedback loops. The result is a new way for Amazon teams to “work backwards” with AI, turning abstract ideas into validated customer insights at scale.

The Preferred Brand Policy is a powerful tool that allows procurement managers to guide their teams toward preferred or restricted brands. However, many users struggle to identify which brands to include due to limited guidance and lack of recommendations. This project introduces brand recommendation ads that intelligently surface relevant and comparable brands, enabling business customers to make informed, confident purchasing decisions while improving discoverability for sellers.