Dataloom

Project Lead: Alexander van Renen

Description

Dataloom is a research project by the Data Systems Lab at the University of Technology Nuremberg (UTN) that explores how large language models (LLMs) and database technology can be combined to make data preparation and analysis accessible to everyone. Our goal is to enable users to turn messy, heterogeneous data into meaningful insights without requiring extensive programming or data engineering expertise.

By combining the flexible reasoning capabilities of LLMs with the efficiency and reliability of traditional data systems, Dataloom orchestrates agentic data pipelines for discovering, cleaning, transforming, loading, and analyzing data. LLMs handle ambiguous, “soft” tasks, while established database techniques are used whenever deterministic and efficient processing is possible.

Dataloom provides a transparent and interactive environment in which users can inspect, refine, and re-execute every step of their data pipeline, minimizing the time from raw data to actionable insights.

Collaborators

• Luca Heller
• Macallyster Edmondson

Code