# KORA Scribe

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Dante Ramirez](https://cerebralvalley.ai/u/danteramirez)
- **GitHub:** https://github.com/dantermz/kora-scribe
- **Demo video:** https://youtu.be/2ZvitbN4au0
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2363

KORA Scribe is a hands-free clinical scribe for public-sector doctors in Mexico, where a primary-care visit lasts 15–20 minutes and much of it goes to typing instead of looking at the patient.

The doctor wears Ray-Ban Meta glasses. An iPhone app records the consultation through the glasses' microphone, Muse Voice Transcribe turns it into a diarized, timestamped transcript, and Muse Glimmer drafts a SOAP note in which every line cites the exact phrase it comes from.

The problem with general AI scribes is that notes look complete but quietly add things nobody said: an inferred diagnosis, a rounded number, a medication the patient mentioned written as if the doctor prescribed it. So before the doctor sees the note, a verifier runs six checks outside the drafter's control: verbatim citations, source and section, scope (Muse Spark Open, a different model from the drafter), coverage, lost audio (unanswered questions, long silences), and a recap check that compares the doctor's spoken summary with what the patient said.

The doctor reviews each colour-coded line next to the transcript, plays back the exact seconds cited, corrects what needs correcting, and signs. Signing is blocked until every red line is decided. The signed record keeps the model's version, every change, and hashes of note and transcript.

Documentation, not decisions: KORA Scribe never diagnoses or recommends treatment. Audio only; the glasses' camera is never used. Synthetic consultations only.

Transparency note: I work as a clinic physician in Mexico's state public health service. My role is seeing patients only: I hold no managerial, director or decision-making position, and I have no say in purchasing, policy or technology. Separately, I am the founder of a health-tech startup, and I entered this hackathon as that founder, not through my clinical job. KORA Scribe was built on my own time and equipment, recorded in a private office, with synthetic data only, and has no connection to my employer. I have asked the organizers to confirm how the "government-affiliated" eligibility rule applies to my case, and I will follow their decision. If this makes me ineligible for prizes, I would still like the project to be judged or shown without competing.

## More from Meta Global AI Developer Hackathon

- [Epigen](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2360)
- [Team Wexy](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2361)
- [Anzo Benjamin](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2362)
- [transformadores](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2364)
- [Pipeline Copilot IanStalter123](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2365)
- [Stuart Johnson](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2366)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2363. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt