# NDxGenius

- **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:** [Nishant Das](https://cerebralvalley.ai/u/QuantHackNDX)
- **GitHub:** https://github.com/NishantDas0079/Evidence-Pilot.git
- **Demo video:** https://youtu.be/zW5vMNmg0gQ
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/707

EvidencePilot checks a coding agent's final report against the run it actually recorded.

Coding agents finish with a summary like "I fixed the bug and all tests pass", and people trust it because reading the full log takes too long. But the agent wrote that summary itself, and it can be wrong: it may claim tests passed when they ran before the last edit, or claim a test run that never happened.

EvidencePilot splits the report into individual claims and judges each one as Supported, Contradicted, or Not settled by the record. Every verdict comes with a quote from the recorded run, and the page can jump to that exact line. Plain code counts the key facts (test runs, last result, files written, whether tests ran after the last edit), and code also checks that every quote really appears in the record, downgrading any verdict whose quote does not.

Both the coding agent and the auditor run on Muse Spark through Meta Model API, using function tools, strict JSON-schema output, and reasoning effort settings. Built this week; no code from earlier prototypes is included.

## More from Meta Global AI Developer Hackathon

- [Theotech](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/704)
- [Intenzo](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/705)
- [Horia BEschea](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/706)
- [ChartCortex](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/708)
- [NinetyNine99](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/709)
- [NexussAI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/710)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/707. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt