# The U-Team

- **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:** [Shriram Vasudevan](https://cerebralvalley.ai/u/shriramkv)
- **GitHub:** https://github.com/giri-rajan/inspectAgent
- **Demo video:** https://youtu.be/04dya3iICFk
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1950

Small foundries and pump makers, like the ones around Coimbatore, still check castings by eye. Machine vision would help, but it needs thousands of labelled defect photos plus expensive integration, so most of these factories never adopt it.

InspectAgent works from the first part with no labelled data. You describe defects in plain words, and the factory’s own acceptance rules decide the verdict.

An operator photographs a part with a phone or line camera. SAM 3.1 segments defects such as porosity, cracks, burrs and scratches straight from text prompts. Muse Spark then checks each region against the part spec and the factory’s accept/reject criteria, rates its severity, and returns a pass or fail with a clear explanation. When the same defect keeps appearing, InspectAgent links it to process data like machine, shift, batch and supplier lot, and drafts an 8D/CAPA root-cause report for the quality engineer. Each defect the engineer confirms is added to a visual library, so inspection gets sharper with use.

The result is consistent, explainable inspection that a small factory can start using on day one, with no data collection project and no costly integration.

## More from Meta Global AI Developer Hackathon

- [Squadroot (Datalent LLC)](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1947)
- [Reflection Lab](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1948)
- [ShiftSense](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1949)
- [Tokiarivelo](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1951)
- [Pranav Paharia](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1952)
- [MindFlayer](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1953)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1950. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt