Life Valuation Actuarial Document Q&A machine using Retrieval Augmented Generation (RAG)
An example retrieval-augmented generation application that lets actuaries ask questions over life valuation actuarial documents, such as statutory accounting, principle-based reserving and IFRS 17, citing the source context behind each answer.
This project builds a RAG pipeline so actuaries can ask natural-language questions across a collection of life valuation actuarial documents, covering statutory accounting principles, principle-based reserving, cash flow and asset adequacy testing, actuarial standards of practice, US GAAP, IFRS 17, and other jurisdiction- and risk-related documents. Because RAG can hallucinate, the tool shows the retrieved context behind each answer so users can verify it.
Documents are chunked at ingestion with langchain-text-splitters and embedded into a vector database. At query time the retrieval pipeline calls the Pinecone, Voyage, Cohere and Anthropic SDKs directly, without a LangChain orchestration layer in the request path, for low latency, and an Anthropic Claude model with a long context window generates answers that cite across multiple retrieved sources.
Built with Python and Streamlit, with a public demo app deployed on Streamlit. Apache-2.0 licensed.