This research presents novel techniques to drastically reduce costs and processing time for large-scale LLM data analytics workloads.

• Addresses critical efficiency challenges where processing 15GB of data can cost $10K and take a full day
• Introduces optimization strategies for batching similar queries and eliminating redundant processing
• Presents a comprehensive framework for making LLM analytics economically viable at scale
• Enables organizations to leverage natural language capabilities across large datasets without prohibitive costs

For engineering teams, this research provides practical pathways to implement LLM-powered analytics solutions that are both performant and cost-efficient, potentially transforming how organizations extract insights from unstructured data.

Optimizing LLM Queries in Relational Data Analytics Workloads