The System Design Methodology of an Interactive Big Data Platform Based on Universal LLMs

This work elaborates on the design and implementation of an interactive big data platform system based on a universal LLM, including the main technical implementation plans and module designs of the system. The system employs a stepwise Text-To-SQL approach, utilizing few-shot techniques, RAG, and multi-source heterogeneous data access technologies, which can significantly enhance the direct executability of SQL. Based on this research, we have upgraded the original China Mobile Smart University Big Data Platform, empowered by AI to provide interactive data querying capabilities. This addresses the high professional barriers in data querying and analysis, and the weak front-end user perception of traditional big data platforms. It enables more people to harness the power of data.