# DataBridge AI

- **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:** [Alex Elaryan](https://cerebralvalley.ai/u/AlexElaryan)
- **GitHub:** https://github.com/AlexElaryan/DataBridge-AI.git
- **Demo video:** https://drive.google.com/file/d/1YJcWNBs7SbGanp-xUC_3Wemip1xH9d3T/view
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2276

DataBridge AI is a human-in-the-loop supplier data onboarding tool for retailers and marketplaces. Supplier product feeds often arrive as CSV files with inconsistent headers, formats, missing values, duplicate identifiers, and invalid barcodes. Teams end up manually mapping fields and investigating import issues.

The application profiles an uploaded CSV, then uses Meta Muse Glimmer to propose mappings from supplier-specific columns to a canonical six-field schema: SKU, name, barcode, price, stock, and category. Local Python heuristics independently cross-check the suggested mappings, and a human approves them before any transformation. Python then normalizes supported values and runs deterministic checks for missing required fields, numeric parsing, duplicate identifiers, and GTIN check digits.

A bounded diagnostic agent can select up to three tools from a fixed allow-list. Python validates and executes those tools; Muse Glimmer can summarize aggregate findings into prioritized remediation guidance. The final readiness state—BLOCKED, REVIEW REQUIRED, or READY TO IMPORT—is decided by Python, never by the model.

The operator can export normalized CSV data, issue and transformation reports, an Excel report, a JSON Schema contract, and a ZIP audit bundle with run metadata and mapping provenance.

The prototype is intentionally scoped to one supplier product-catalog CSV and the six-field schema. It does not currently ingest JSON/XML, join multiple feeds, write directly to production PIM/ERP systems, or send supplier messages automatically. DataBridge focuses on reducing manual onboarding work while preserving human approval, deterministic validation, and an auditable trail.

## More from Meta Global AI Developer Hackathon

- [Uplift](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2273)
- [Saad](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2274)
- [ZeroHP](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2275)
- [Cactus Sentinel](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2277)
- [KLI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2278)
- [WEB NOVAS](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2279)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2276. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt