Extract structured metadata from messy sources - PDFs, images, free text, tables, and more
Handle any number of metadata fields, from study design to sample attributes, with ontology-mapped outputs
Automate what used to take weeks - Polly Xtract delivers 4× faster curation with zero missing fields
Outperforms GPT-4 and even manual curators in accuracy, F1 score, and consistency
Outputs come with field-level evidence, confidence scores, and explainable reasoning logs
Already powering real-world workflows - from GEO harmonization to clinical trial parsing and EMR digitization
Schema-flexible by design - works with both standard models (like CellxGene) and your internal data formats