# Durre Shehwar

- **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:** [Durre Shehwar Khan](https://cerebralvalley.ai/u/KDS)
- **GitHub:** https://github.com/Durreshehwar15/TraceMind/
- **Demo video:** https://youtu.be/KPUQR39QROk
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3193

TraceMind – From Bugs to Verified Fixes

Software debugging is one of the most time-consuming challenges developers face. A runtime error may appear in one file while its actual cause lies in another. Developers must manually investigate stack traces, understand dependencies, identify the root cause, implement corrections, and test whether those corrections introduce new problems. Although AI coding assistants can suggest fixes, generated code is not necessarily correct or safe to apply.

TraceMind is an AI-powered root cause analysis and automated debugging verification platform designed to bridge the gap between identifying a bug and proving that its proposed fix works.**

Powered by Meta's Muse Spark Open model through Baseten, TraceMind analyzes runtime errors, stack traces, and multi-file project context to identify potential root causes and generate targeted code patches. Instead of blindly modifying source code, it provides developers with patch previews, code diffs, and SHA-256 file integrity checks to detect outdated or modified files before applying changes.

TraceMind combines Docker-based isolated bug reproduction with Modal-powered remote regression testing. Once a developer reviews and applies a proposed patch, AI-generated regression tests can be executed in a separate Modal sandbox to evaluate whether the correction behaves as expected. Developers receive actual execution output and test results rather than relying solely on AI-generated explanations.

What makes TraceMind different is its verification-first approach to AI-assisted software debugging. It does not treat an AI-generated patch as a guaranteed solution. It connects error reproduction, root cause investigation, patch generation, safe review, and execution-based verification into one workflow.

The core problem TraceMind solves is the gap between an AI suggesting a code fix and a developer knowing whether that fix actually works.

Its vision is to make debugging more explainable, reliable, and evidence-driven, helping developers spend less time chasing errors and more time building dependable software.

TraceMind: Don't just fix the bug. Prove the fix.

## More from Meta Global AI Developer Hackathon

- [Team OwoBook](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3190)
- [Reality OS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3191)
- [CropAdvisor](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3192)
- [Motion Clone Team](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3194)
- [Vibe Coder's Life](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3195)
- [Sidethought](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3196)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/3193. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt