# TitleGuard

- **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:** [Mohamed A](https://cerebralvalley.ai/u/adib0524)
- **GitHub:** https://gitlab.com/ess-cor/titleguard
- **Demo video:** https://huggingface.co/spaces/esscor/titleguard/blob/main/TitleGuard_Demo.mp4
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2240

Agentic land-title fraud checks for Tamil Nadu property documents, built on Meta Muse Spark.

Buying land in Tamil Nadu means trusting a stack of part-Tamil, part-English papers: the Patta, Chitta, Encumbrance Certificate (EC) and sale deed. Fraud hides in the gaps between them: a seller who is not on the patta, a co-owner left out, a deed selling survey 45/3B when the patta holds 45/3A, 40 cents sold on a 30-cent patta, or a bank mortgage never released. Today only a lawyer reading every page catches this, so most small buyers skip it. TitleGuard reads the documents with Muse Spark vision and structured output (Tamil and English to JSON), then a Muse Spark tool-calling agent cross-checks every field across documents using five deterministic rule tools (EC coverage, owner names with Tamil transliteration and co-owner check, survey numbers, extents across hectare/acre/cent/sq.ft, unreleased encumbrances). Findings only come from tools, so the model cannot invent a fraud flag. The result is a risk report with red flags, why they matter, what to ask the seller, and the full agent trace. The Meta key stays server-side. A labelled demo mode runs the same agent on four fictional Madurai-area cases. Built by a Madurai real-estate business owner for the buyers he meets every week.

Live demo: https://huggingface.co/spaces/esscor/titleguardnam

## More from Meta Global AI Developer Hackathon

- [AugentFlow](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2237)
- [Phantom Labs](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2238)
- [In Character](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2239)
- [Lost & Frame](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2241)
- [Moria Crew](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2242)
- [Tygart Media](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2243)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2240. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt