400 sources synthesized across web and academia.

To automate this process, I built the research pipeline for the India market as part of a consulting project through Utrecht University's Center for Global Challenges and the Deep Transitions Lab, working within a cross-disciplinary team of professionals.

The system used a multi-agent researcher → writer → polisher architecture. A researcher agent gathered evidence from academic and web sources: academic search used Semantic Scholar, retrieved full papers, and used RAG to retrieve and synthesize evidence from them, while a custom SearxNG deployment was used by the agent to gather evidence from the web. A writer agent synthesized the findings into a structured report with citations, and a polisher agent refined the final output for clarity and consistency while preserving the source trail. The resulting reports were added to a shared knowledge base used by other parts of the project. Every research output retained citations back to its underlying sources.