Inspiration
Brandon and I walked into a room of 60 people knowing just one other person. We thought how could we seize the opportunity to network more efficiently despite knowing no one. That's where AI could help us.
What it does
RoomPilot is an iOS event companion that shows you who's attending, where they're gathered, and how you're connected. Making leads and conversations easier and more personal. Personal AI features like connection recommendations and conversation context stay entirely on-device and private.
How we built it
The iOS app is built in SwiftUI. The backend runs on Cloudflare Workers with a SQLite database, with Sign in with Apple authentication and on-demand Apple Wallet pass signing. On-device AI handles personal context. It leverages SFSpeechRecognizer and Foundation Models to bring those connections to life.
Challenges we ran into
Narrowing down scope for this was a challenge. The potential for features in this are endless as are real human connections. Building a fully hosted edge backend along with the on-device logic was a real time stretcher but making certain trade offs were especially challenging.
Accomplishments that we're proud of
We are proud of the amount of features we were able to integrate, and the teamwork we had with efficiently splitting up tasks. The graph on the Home Screen we feel really sells the idea and for that we are proud.
What we learned
On-device and cloud AI aren't competing they're complementary layers with different jobs. The hardest decisions were about what not to build on the server.
What's next for RoomPilot
Bluetooth proximity for automatic cluster detection, AI-powered conversation starters combining shared graph with on-device context, multi-event professional identity, and organiser analytics for real-time crowd flow and a whole lot more.
Built With
- cloudflare
- core-data
- coreml
- foundation-models
- swiftui
- vision
Log in or sign up for Devpost to join the conversation.