A phone and tablet recipe app built around the annoyances of cooking: messy hands, jargon you keep looking up, paper recipes you cannot use. A wake word starts voice control that answers instantly from a local model, and questions about terms or substitutions are handled with RAG. Photograph a magazine page and it becomes data.
Judges called voice-only operation the most direct and clever answer to hands busy with cooking. They praised the use of Gemma 3n with speech recognition, and that it answers sudden questions like ingredient substitutions, not just recipe steps.
As the name of the award suggests, the decisive question was probably where to put the small model. Voice control, where response time is the experience, runs locally; questions that need knowledge are routed to RAG in the cloud. That split reads as a design made by someone who understands what each model is good at. Photographing paper recipes matters too, since it lets people keep what is already in their kitchen and lowers the friction of adopting the app at all.
See all winners of 第2回 AI Agent Hackathon with Google Cloud →
Summaries are written by this site. Project pages include demo videos, screenshots and the team's own write-up (videos may autoplay).