# Handover

- **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:** [Siddharth Prabhakar](https://cerebralvalley.ai/u/Wakeupsid)
- **GitHub:** https://github.com/Siddharthprabhakar/Handover
- **Demo video:** https://youtu.be/5ULRvvWC-4g
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1263

The problem. University research labs, especially in HCI and the social sciences, depend on students who rotate every year or two. What they know lives in chat threads, Drive folders, consent forms and their heads. Quiet students get stuck without saying so, assistants get overloaded with no visible limit, and when someone leaves, the next person starts over.

What Handover does. Each week, students answer four short questions through a personal link, typed or spoken. Meta Voice Transcribe turns speech into text, and Muse (muse-spark-1.3) turns the answers into a digest for the professor: who is stuck, in their own words, who needs tasks and who has not checked in. Interviews are recorded under a consent gate and transcribed with speaker labels. Muse finds themes with exact quotes in each interview, then across interviews for each research question, and suggests which study a recording belongs to. When a student leaves, Muse drafts questions from their own tasks and a thank-you note from what they finished. The next person gets a brief and a study pack with research questions, IRB status, participant progress and evidence.

AI proposes, code enforces. A quote is kept only if it appears word for word in the transcript. Participant counts and the workload limits (5 tasks and 10 hours a week, overtime with a reason, a recovery week, a 20-hour hard cap) are plain code with tests. A person approves everything Muse suggests.

Built solo during the hackathon with Next.js, Supabase and the Meta Model API. All ten Muse jobs and speech-to-text run on Meta.

## More from Meta Global AI Developer Hackathon

- [HereOut](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1260)
- [Baris Erzeren](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1261)
- [A Better U](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1262)
- [Brak](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1264)
- [bAIwor](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1265)
- [Truth Layer - Exauce Kilonga](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1266)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1263. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt