Upload a medical PDF and our AI pipeline handles the rest — parsing documents, extracting clinical data, standardising results, and generating actionable health reports.
Every medical record passes through a multi-stage AI pipeline — each stage purpose-built to extract, structure, and enrich clinical data with maximum accuracy.
Medical records uploaded as PDFs
Document structure analysed page by page
Biomarkers, diagnoses, procedures, genetics
Standardised names, units, and clinical context
AI-generated clinical health reports
Upload any medical PDF — lab reports, clinical notes, imaging results, genetic tests, or discharge summaries. Our platform accepts records from any healthcare provider, in any format.
No standardised template required. Upload PDFs from hospitals, labs, clinics, or specialists worldwide.
Records are stored in encrypted cloud storage with signed URLs. Access is scoped to the patient and their clinician.
Upload individual records or entire patient histories at once. Each document is processed independently through the pipeline.
Before extraction begins, each document is parsed and every page is intelligently classified. This pre-screening step identifies which pages contain clinical data and what type — so only relevant content reaches the extraction models.
Advanced OCR and visual language models extract text, tables, and images from each page — even handling scanned documents, handwritten notes, and complex table layouts.
Each page is classified by data type — biomarker results, clinical findings, genetic data, or non-medical content. Pages without clinical relevance are filtered out early.
Pages with visual content (charts, scanned tables) are routed to specialised vision models, while text-heavy pages use faster text models — optimising both accuracy and speed.
Four specialised AI processors run in parallel — each trained to extract a specific category of clinical data with high precision. Every data point is attributed to its source page and validated against clinical standards.
All four data categories are extracted simultaneously across multiple pages, with concurrent processing for maximum throughput.
When test dates are missing from a page, the system propagates dates from adjacent pages and cross-references between data types to ensure temporal accuracy.
Duplicate entries that appear across multiple pages are automatically detected and merged, preserving the most complete version of each data point.
Different labs use different names and units for the same test. The enrichment stage solves this by canonicalising biomarker names, standardising units, and adding clinical context — making data comparable across providers and over time.
Medical embedding models cluster similar biomarker names (e.g. "HbA1c", "Hemoglobin A1C", "Glycated Haemoglobin") under a single canonical name — enabling longitudinal tracking across providers.
Results are converted to standard units so values from different labs are directly comparable. Reference ranges are recalculated to match, preserving clinical meaning.
Each biomarker is assigned to health areas (cardiovascular, metabolic, thyroid, etc.) and clinical groups — enabling filtering, organisation, and targeted analysis by body system.
With structured, standardised data in place, AI generates comprehensive health reports tailored to each patient. Multiple report types are available — each designed for a different clinical lens and consultation need.
Every insight and recommendation is grounded in the patient's actual data and peer-reviewed clinical guidelines. The AI cites its sources and explains its reasoning.
From comprehensive clinical analysis to focused biomarker trend tracking — choose the report type that fits the consultation. Each offers a distinct clinical perspective.
Reports surface the insights that matter most — flagged values, emerging trends, risk factors, and prioritised recommendations — ready for clinical decision-making.