# Cadence

- **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:** [Harshal Vaidya](https://cerebralvalley.ai/u/try_catch_me)
- **GitHub:** https://github.com/theloopsmith/Cadence
- **Demo video:** https://www.youtube.com/watch?v=SCIVcCh8lok&feature=youtu.be
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1466

Cadence is an iPhone companion paired with Ray-Ban Meta glasses, built to support everyday independence for people living with Parkinson’s. It combines Meta’s language, vision and speech models with on-device motion processing to make everyday assistance easier to access.

Three Meta models work together across the experience.
- Muse Spark Open interprets conversational requests, proposes concepts to look for in photos, and checks handwriting-photo quality so users can retake unclear images.
- SAM 3.1 independently locates those concepts in photos captured through the glasses. Cadence uses the grounding results to provide tentative object descriptions and explain incomplete results.
- Muse Voice Transcribe provides online speech transcription, helping turn spoken requests into actions within the companion.
This coordination connects a user’s question to visual evidence and a bounded response. Judges can trigger each integration through dedicated website tiles and inspect live model results using fixed samples.

Cadence’s guided Daily Check brings together a short voice sample, a handwritten sentence captured through the glasses, and a brief walking sample. Spoken instructions guide each step, and users can pause, skip or stop. The resulting observations create a record users can review over time.

Beyond these checks, Cadence supports routine reminders, medication and meal-time logging, and rhythmic audio cues for possible walking pauses. Motion processing runs locally, while durable offline queues preserve records through connection interruptions.

Cadence has been tested with physical Meta glasses, including real photo capture, microphone input, spoken replies, walking cues, reconnects and background operation. It is a daily-living and wellness prototype that records observations without providing diagnosis or treatment advice. Caregiver escalation is explicitly simulated in the public demo.

## More from Meta Global AI Developer Hackathon

- [ekatra maya](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1463)
- [famparty.com | Party with your fam!](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1464)
- [iVibe](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1465)
- [KEVIQLABS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1467)
- [Confio](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1468)
- [Orgenix](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1469)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1466. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt