Mixpeek extracts entities from PDFs, medical imaging (DICOM), scanned documents (OCR), FHIR bundles, and lab data, then matches against eligibility criteria with source attribution to the exact document and finding.