Lessons Learned Building a GenAI Powered App
summary
A GenAI quiz app called Quic generates quizzes and images using Gemini Ultra and image‑gen v2, hosts UI with Flutter and Cloud Run and stores state in real‑time Firestore, and serves 100+ live players in a demo. Prototype takes ~seven weeks, generator plus validator yields about 91–94% accuracy on benchmarks, but models are inconsistent, slow and hallucinate. Engineering responses include pinning model versions, defensive parsing of JSON, background validation by an LLM-based assessor, and caching/parallel calls to mask latency.
AI-written summary · about
chapters (9)
- 0:00GenAI trivia app journey
- 1:49Weekend project origins
- 18:32Architecture and tech stack
- 25:18Static to AI content
- 28:50GenAI challenges
- 31:43Lessons from GenAI
- 43:50Testing AI content
- 57:19Summary and future
- 59:52Q&A on filters and temp
related talks
46mLessons Learned Building A GenAI Powered App
03mAI generated trivia games on Cloud
01h 11mBuild Your Own AI Sidekick with Flutter
01h 03mBuild & deploy AI-powered apps
38mGenerative UIs: Using LLMs to Create Flutter Screens at Runtime
30mGenAI on Android with Gemini Nano and Firebase
39mGetting Started with GenAI for Android Apps
23mScaling GenAI Applications and Agentic AI for Business Competitive Differentiation