Nora

Inspiration

Every year, tens of millions of Americans visit emergency rooms for conditions that could be safely handled at an urgent care clinic, a telehealth call, or even at home, driving up costs and clogging hospitals for people who truly need them. We kept coming back to that anxious question: should I actually go to the ER, or will I be fine? There's no good tool that answers that question quickly, privately, and without requiring you to wade through generic WebMD articles. So we built one.

What it does

Nora lets you speak or type your symptoms and receive a clear, actionable recommendation in under 60 seconds: self-care, telehealth, urgent care, or emergency room. Nora asks only the follow-up questions that actually matter (typically 1 to 3), can optionally analyze a photo of a situation, and routes around the conversation entirely if it detects a life-threatening emergency like chest pain or stroke symptoms. Everything runs entirely on your device with no account, no cloud upload, and no waiting.

How we built it

We used Android Studio and Kotlin to create a working Android app

Challenges we ran into

Getting reliable, structured outputs from a medical LLM within a strict token budget, while also streaming tokens to the UI for perceived responsiveness, required a lot of work.

Accomplishments that we're proud of

We're proud that Nora handles the full spectrum of triage, from a sore throat that needs rest to a chest-pain emergency that bypasses the AI entirely and routes straight to 911 in under a second

What we learned

We learned that on-device ML is both more capable and more brittle than cloud inference, since model quantization and prompt token budgets all become your problem in ways that an API call abstracts away.

What's next for Nora

We will be trying to deploy this app in the real world to get users.

Built With

• kotlin