Generative AI Music Platform

AiMi is a generative music platform. Instead of static three-minute songs, it produces continuous musical "experiences" that listeners can shape in real time. I joined as founding Product Manager and led the app from concept through production launch, years before generative AI became a mainstream category.

Beyond the core product, I developed AiMi's first paid partnerships with two national retail chains and helped scale the consumer app to 600K+ users.

Generative AI music didn't map cleanly onto existing listening habits, so I ran more than 100 user interviews to understand how it actually fit into people's lives (background focus music, workouts, mood regulation) and synthesized that into use cases and journey maps that prioritized what the product should do first.

The core design challenge was translating a complex ML system into something a casual listener could use without thinking about it. We built human-in-the-loop feedback (thumbs up/down, shuffle, and direct control over individual elements like harmony, bass, and melody) so listeners could steer the generative output in real time instead of just passively hearing whatever the model produced. That sense of control turned out to be a meaningful driver of sustained retention.

None of this worked without close, direct collaboration with the ML research team: dataset inputs, evaluation criteria, quality metrics, and latency constraints. Product and research had to shape each other, since what listeners needed informed what the model was optimized for, and what the model could realistically do in real time shaped what the product could promise.