Prometheus is a web-based platform that aims to streamline the process of uncovering insights within data by generating and evaluating concrete analysis tasks. It is powered by an LLM, which analyzes variables in a dataset and outlines initial tasks unique to that dataset. Prometheus generates corresponding code snippets that evaluate the generated tasks, enabling a researcher to quickly iterate through simple exploratory analysis and identify more complex insights. The platform is designed to augment data analysis rather than replace it, and its user interface allows for interactive reprioritization, editing, and expansion of generated tasks.
Omar Shaikh, PhD Candidate, Stanford University; Caleb Ziems, PhD Candidate, Columbia University