MediScribe

Landing Page
Recording Page
Conversations List
Transcribed Dialog Page
Summary Page

Inspiration

Ever leave a doctor’s appointment and forget key details? Or see a new doctor and struggle to recall what the previous one said? MediScribe solves that gap by capturing and summarizing visits so patients and providers can revisit the important information anytime.

What it Does

Records doctor–patient conversations
Transcribes speech and labels speakers
Generates a structured medical summary
Lets users ask questions about the visit with AI

How We Built It

Frontend: Next.js (React + Tailwind)
Backend: Node.js + Express
Transcription: ElevenLabs Speech-to-Text
Data & Files: Firebase (Firestore + Storage)
AI Summaries & Chat: Backboard + Gemini

Challenges We Ran Into

One of our biggest challenges was orchestrating multiple services such as speech-to-text, databases, file storage, and AI assistants while making them work smoothly. We also had to pick up new technologies like Firebase, Backboard and LLM API's on the fly, which came with a steep learning curve. We also had figure out speaker diarization since identifying different speakers didn’t automatically tell us who was the doctor or the patient.

Accomplishments We’re Proud Of

As this was our first and second hackathons, we’re proud that we delivered a full end-to-end workflow: record → transcribe → summarize → view, using real conversation data. We successfully generated AI-powered medical summaries and enabled Q&A over the visit and by the end of the hackathon, we had a working prototype with persistent data.

What We Learned

We gained hands-on experience integrating LLMs with transcription systems, handling uploads and structured outputs, and designing summaries with clinical clarity in mind, all while learning new tools under tight time constraints.

What’s Next

Next, we plan to add authentication and login, support persistent chat threads per visit, allow users to verify speaker roles, and enable exportable summaries for EMR sharing.