We are entering an era where AI agents don't just generate text. They build the interface itself. Frameworks like Flutter’s Generative UI SDK allow apps to assemble screens dynamically at runtime, creating highly personalized user experiences. But this power comes with a massive risk: when the UI is no longer hard-coded, a single malformed stream or "hallucinated" widget schema can shatter the entire application instantly, leaving users with a broken screen. This roundtable is a reality check for developers building these dynamic, agent-driven interfaces. We will move beyond the "happy path" demos to discuss the hard operational challenges: How do you build "safety cages" that validate AI output before it renders? How do you handle fallback states when the model gets it wrong? Join us to share your war stories and strategies for keeping dynamic UIs stable in production.

Shipping a Flutter demo is easy. Shipping a system that survives real-world scale, messy backend constraints, async race conditions, QR-based access control, and 2 lakh users before a festival deadline? That’s different. In this talk, I’ll break down how to design Flutter apps like systems engineers — not just UI developers. Using real production case studies (large-scale QR-based access management + Firebase-backed systems), we’ll explore how to design resilient, scalable Flutter architectures from day one — without over engineering. This session is for developers who want to move from “app developer” to “product architect.” Attendees will learn: -How to structure Flutter apps using scalable architecture (MVVM + BLoC hybrid patterns) -Designing for async-heavy systems (QR scanning, network latency, offline states) -Production-ready Firebase integration strategies (caching, batching, cost control) -Preventing performance bottlenecks before scale hits -How to think like a CTO when building Flutter products