The Scenario
A global fintech company is building a customer support assistant. The assistant retrieves data from a vector database of help documents. However, raw search results can contain irrelevant or outdated information, causing hallucinations. You need to build a comprehensive RAG prompting workflow that evaluates retrieved document relevance, reranks documents, runs compliance filters, and generates a grounded response.
The Brief
Create an advanced prompt system that orchestrates RAG context ingestion, document relevance filtering, cross-document consolidation, and final grounded answer generation with inline citations.
Deliverables
- A complete orchestration prompt template containing system instructions, context variables, formatting blocks, and guardrail rules.
- A document-filtering validation instruction to classify retrieved documents as "relevant" or "irrelevant" based on the user question.
- A detailed hallucination mitigation prompt strategy that forces the LLM to only answer using retrieved facts and cite sources.
- An optimization analysis proposing token cost management, handling larger document payloads, and mitigating prompt drift.
Submission Guidance
Document the complete prompting architecture in Markdown. Provide detailed system instructions and explain the formatting variables and guardrails with code examples.
Submit Your Work
Your submission is graded against the rubric on the right. If you pass, you get a public Badge URL you can share on LinkedIn. There is no draft save, so work offline first and paste your finished response here.