Research as a Code - final

Compare NVIDIA Universal Deep Research, and Google Test-Time Diffusion Deep Researcher side-by-side * three chunked and indexed data collections from specific domains: congress (Congressional legislation data), us_tariff (US Tariff code documents), sustainability (SDG - Sustainable Development Goals and other sustainability documents) * Milvus Vector DB capable of hybrid search, Milvus is the Vector DB also used by NVIDIA RAG NIM * Also based on the AI-Q NVIDIA Research Assistant blueprint and the NVIDIA RAG Blueprint * Deployable to AWS, uses NIMs * IaaC leveraging Karpenter on EKS for frigal scale-down and quickest scale-up * CopilotKit AG-UI and custom streaming for execution details Source code: https://github.com/Research-as-a-Code/Research_as_a_Code Also see: https://devpost.com/software/research-as-a-code