To build fast and cost-effective web applications, developers can split AI workloads between the user's browser and the cloud using Firebase. This talk demonstrates this hybrid approach through a live web app demo: we use a lightweight, in-browser model (Gemini Nano) to quickly handle simple client-side tasks like generating image descriptions and alt text, while routing complex demands—like live web searches and image generation—to powerful cloud-based models like Gemini Flash. Attendees will learn how to implement this practical workflow to balance performance, cost, and user experience in their own web projects.