Altar: Structuring Sharable Experimental Data from Early Exploration to Publication

A team of researchers including William Gaultier, Andrea Lodetti, and Aliénor Lahlou has introduced Altar, a novel framework designed to solve the persistent problem of managing experimental data during the active development phase of research projects. Published on arXiv (2602.18588), Altar addresses a critical gap in data management plans that typically focus on final publication rather than the messy, iterative process of exploration.

The framework is built around the established Sacred experiment-tracking model and employs a hybrid storage architecture. Parameters, metadata, curves, and small files are stored in a flexible NoSQL database, while large raw data files remain in dedicated storage systems, linked through unique identifiers. This approach maintains efficiency and traceability without imposing rigid data models that could hinder early-stage research creativity.

Altar is designed to be composable with existing workflows, requiring minimal disruption to research habits. The team has documented multiple usage pathways tailored to different skill levels, from PhD students to laboratory administrators. While the system can be used without specialized infrastructure for getting started, it can scale to server deployment for public accessibility when preparing data for publication. By focusing on the dynamic phase of research that's often overlooked, Altar provides a practical bridge between exploratory experimentation and FAIR-aligned data sharing, potentially transforming how collaborative research projects manage their most valuable asset: their data.