# Ajali Kit

- **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:** [Ian Odhiambo](https://cerebralvalley.ai/u/Ianodad)
- **GitHub:** https://github.com/Ianodad/ajali-kit
- **Demo video:** https://github.com/Ianodad/ajali-kit/releases/download/demo-video/ajali-kit-demo-3min.mp4
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5340

Ajali is Swahili for "accident". Ajali Kit is a mobile web app (an installable PWA) that helps a Kenyan motorist build a complete motor insurance claim pack at the roadside, in the stressful minutes after a crash.

The problem: After an accident, most drivers don't know what an insurer will ask for. They take a few blurry photos, forget the other driver's licence or insurance details, skip witness contacts and the police OB number, and later write an account from memory. Claims then stall while the insurer asks for missing items, and a weak account or missing evidence can hurt the claim.

What it does: Ajali Kit walks the driver through a short, guided checklist:
- A safety check first, then the incident basics (time, place, what happened).
- Damage photos. Muse Spark finds each damaged area and names the part, damage type and an indicative severity. SAM 3.1 turns each area into a pixel-accurate coloured mask and outlines the whole vehicle, so the app can show roughly how much of the visible car is damaged. The driver confirms, edits or adds any missed damage.
- Document photos (licence, logbook, insurance certificate) are read into form fields with a confidence level, and the driver confirms each one.
- Other party, police and witness details, with consent before a witness's phone number goes in the pack.
- A drafted written account, damage schedule and missing-items checklist, which the driver reviews and confirms in their own words.
- One tap builds a claim pack PDF with annotated damage photos next to the originals, all the details, and a checklist of anything still outstanding. It can be shared, downloaded or printed.

Privacy and safety by design: all data stays on the phone in local storage, and the server routes are stateless. The app never names insurers, never estimates repair costs and never decides fault. AI marks are clearly labelled as a guide, and the driver confirms everything before it goes in the pack.

## More from Meta Global AI Developer Hackathon

- [UniPass AI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5337)
- [Pitwall](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5338)
- [concat](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5339)
- [Ghost](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5341)
- [Burgeoning](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5342)
- [spArkHAT](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5343)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/5340. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt