Adding cloud-based AI to a Flutter app is exciting, but monitoring it in production is a massive headache. Unlike standard APIs that give predictable success or error codes, AI models are completely unpredictable. The core tension lies in tracking a system that changes constantly: two users can ask the same question, but the app might use vastly different amounts of data, run at different speeds, and cost different amounts of money each time, making it incredibly hard to spot real performance bottlenecks before your API bill spikes. This roundtable is a practical space for developers to discuss how they keep an eye on AI features after they go live. We will step away from basic app analytics to look at the broad challenge of tracking AI telemetry, managing usage tokens, and measuring the actual quality of AI responses. Join us to share what tools and metrics you use to keep your AI-powered apps fast, reliable, and affordable.

A few years ago, embedded development meant flashing Android Things onto an NXP Pico board and wiring up Google Assistant to control an LED. That platform is gone, but the problem — building real, controllable smart-home hardware without needing a background in embedded systems — hasn't. Flutter and Raspberry Pi are a surprisingly capable answer today. This talk documents building an actual smart-home app with Flutter running on a Raspberry Pi: setting up Flutter for embedded Linux, building a touch-friendly UI for local device control, wiring up GPIO to control real hardware (lights, sensors), and connecting the app to a backend so the same Flutter codebase can also run as a companion mobile app. Along the way, I'll compare this approach to the Android Things + Google Assistant stack I used to build a similar project on, and talk about what's genuinely easier now, what's harder, and what embedded Flutter still can't do well. Expect a live, on-stage demo of a physical light being controlled from a Flutter UI — built specifically for this talk. Key Takeaways: 1. How to get Flutter running on a Raspberry Pi for embedded/kiosk-style UI 2. Practical patterns for talking to GPIO and real hardware (lights, sensors) from a Flutter app 3. How to structure a shared codebase so the same app logic drives both an embedded touch UI and a companion mobile app 4. A direct comparison of building smart-home hardware with Android Things vs. Flutter on Raspberry Pi today — what improved, what got harder 5. Live hardware demo: controlling a physical light from a Flutter UI on-stage