Regulatory guidelines assistant

The brief

Regulatory and quality teams search long guidelines for one requirement, and a wrong answer can delay an approval. Build an assistant over CDSCO guidance, ICH guidelines and sample SOPs that answers with the exact section cited, and says when the documents do not answer. Good looks like: a pipeline for long guideline PDFs, retrieval compared on 50+ test questions, citations on every answer, and refusals when unsure.

What you will build

1. Collect CDSCO guidance and ICH guidelines
2. Parse and chunk by section with metadata
3. Build hybrid retrieval with re-ranking
4. Write 50+ test questions with the right section
5. Evaluate faithfulness and retrieval recall
6. Deploy with citations and an "I don't know" path

Free, open data to build it with

• ICH guidelines ↗ ICH
• CDSCO ↗ Central Drugs Standard Control Organisation

Check each dataset's licence before you publish your work. Your mentor gives your team its own version of the brief (data, scope or region), so every team's project is its own.