AI/ML Approaches to Automate Scientific Data Curation

The AI for Curation project integrates advanced Large Language Models (LLMs) into the data curation workflow to tackle the growing challenges of managing NASA's space biology datasets; data ingestion, standardization, and user interaction with the curated datasets are a large part of our initiative. The project implements AI/ML approaches to convert unstructured data into standardized formats, significantly reducing the time and effort required for manual curation. Hence, reliable datasets are established through automated processes such as data validation and verification, while curators can focus on scientific analysis. Our work supports NASA's mission to optimize workflows and promote open science, guaranteeing that the knowledge and resources generated are accessible and valuable to the scientific community as a whole. Primarily, our AI-Cure tool will provide a secure web-based interface to upload scientific files, generate AI summaries and keywords, and interact with content through a Retrieval-Augmented Generation (RAG) chat interface. The conversational agent leverages a growing repository of curated studies from the Open Science Data Repository (OSDR) to help users navigate and retrieve relevant information. We are actively expanding the tool’s capabilities to support multimodal content, including images, PDFs, PowerPoint presentations, and Excel files. These developments enhance the platform’s versatility and usability, supporting researchers across disciplines. Ultimately, this work contributes to NASA’s broader goals of promoting open, reproducible science and creating reliable processes for managing scientific knowledge.