Lumina

AWS database services support millions of customers running mission-critical workloads. Despite collecting thousands of feedback items through the AWS Console, teams struggled to extract meaningful patterns and insights from this wealth of information, creating a growing disconnect between customer needs and product decisions.

The challenge went beyond just volume. The technical complexity of database services made it difficult for designers and product teams to understand feedback without deep domain expertise. Teams faced a massive backlog of customer comments spanning different stages of product maturity, with no clear way to prioritize which issues to address first.

When product teams are focused on technical implementation details, it's easy to lose sight of the actual human problems customers are trying to solve. We needed a way to transform raw feedback into actionable insights that could drive customer-centered decisions at scale.

As the designer and developer of Lumina, I led this project from concept to implementation:

• Concept and research: Identified the opportunity to leverage AI for feedback analysis through my firsthand experience with existing feedback processes and discussions with my manager about improving our design workflow through AI
• Design and architecture: Created the multi-dimensional classification system and designed both the backend processing pipeline and frontend visualization
• Technical implementation: Built the entire system using AWS serverless technologies and AI services, from initial prototype to production solution
• Adoption and iteration: Led adoption across AWS Databases and Analytics teams and evolved the system based on ongoing feedback

Lumina transformed scattered customer feedback into structured insights through a multi-dimensional AI classification system that analyzes intent, technical areas, product features, and user journey stages simultaneously.

The system processes thousands of feedback items in minutes instead of weeks, enabling teams to analyze 100% of customer feedback rather than the previous 15-20% sample. This comprehensive view provides a foundation for more informed product decisions.

While direct attribution is challenging, Lumina has contributed to numerous product improvements by highlighting patterns in customer feedback and providing evidence to support design decisions. Its most significant impact has been changing how teams approach customer feedback—transforming conversations from "Do we have time to read feedback?" to "What are we learning from our customers?"