Welcome to the Brighthive AI Agent Engineer challenge. You'll simulate a SaaS platform that integrates with a metadata catalog and a warehouse, then build a multi-agent workflow using LangGraph to extract, validate, and summarize metadata issues.
- Simulate a SaaS integration layer (MCP + warehouse).
- Build 3 LangGraph agents: MetadataAgent, ValidationAgent, SummaryAgent.
- Expose a FastAPI endpoint to run the workflow and return a structured report.
- Install dependencies (in a virtual env or conda):
pip install -r requirements.txt- Run the API server:
uvicorn saas_layer.service:app --reload- POST to
http://localhost:8000/run-workflow
curl -X POST http://localhost:8000/run-workflow| Area | What We're Looking For |
|---|---|
| Correctness | Output is logically accurate |
| Modular Agent Design | Reusable, testable agent logic |
| Integration Thinking | Realistic SaaS/MCP-style structure |
| Code Quality | Clear structure, naming, formatting |
| Initiative (Bonus) | Error handling, testing, extensions |
You should be able to complete this in ~2 hours.
Good luck!