Training LLMs for querying financial data with natural language prompts

The abstract introduces the motivation behind this research: the rapid expansion of financial investment data and the need for Artificial Intelligence-driven solutions for accurate and real-time analysis. The research presents an open-source Large Language Model (LLM) fine-tuned on financial datasets to provide precise investment insights. The LLM integrates MongoDB Atlas Vector Search to store and retrieve vector embeddings efficiently. This allows natural language querying of financial data. The implementation leverages Hugging Face Transformers, Sentence-Transformers, and FastAPI, enabling a scalable framework for real-time financial queries and automated decision-making. The system is deployed as an API, allowing seamless integration into financial platforms. The research contributes to the open-source community by providing a robust AI-powered financial tool for analysts, traders, and researchers.