Our paper "The Case for Instance-Optimized LLMs in OLAP Databases" has been accepted at the 27th International Workshop on Design, Optimization, Languages and Analytical Processing of Big Data (DOLAP 2025)!
DOLAP 2025 is co-located with EDBT/ICDT and will take place on March 25, 2025 in Barcelona, Spain.
The Problem
Large Language Models can augment analytics systems with data summarization, cleaning, and semantic transformation. However, deploying LLMs at scale — processing millions to billions of rows — remains prohibitively expensive in computation and memory.
Our Solution: IOLM-DB
We present IOLM-DB, a novel system that makes LLM-enhanced database queries practical through query-specific model optimization. Rather than using a general-purpose LLM for all queries, IOLM-DB creates specialized, lightweight models tailored to specific analytical tasks.
Key Benefits
- Dramatic reduction in inference costs
- Feasible processing of large-scale datasets
- Maintained semantic capabilities
This work was led by Bardia Mohammadi (MPI-SWS).