Five layouts in. One table out.
Five labs, five different layouts, one schema that has to survive all of them. Same problem as invoices, purchase orders or shipping documents.
- 5 sample reports · View source to inspect
- Fictional patient data
- 100% client-side · no API calls
Pick a sample report
Westside Oral Pathology
LayoutPatient line is a one-liner with age as “47Y”. Diagnosis is buried at the bottom in a red box. Date written as “02-Apr-2026”.
Pacific Dental Imaging
LayoutTwo-column patient block. Findings as bullet list, impression below. Date written long-form (“February 11, 2026”).
Riverside Clinical Lab
LayoutFixed-width Courier table, no special markup — flags as “*L” / “*H” trailing each row. Patient labeled “PT:”, age as “44 yo”.
Northwest Surgical Pathology, P.C.
LayoutThree specimens (A/B/C). Margin status, lymphovascular invasion and pTNM scattered across paragraphs. Date as “06-Mar-2026”.
St. Mary's Medical Center — Emergency Lab
LayoutRed STAT banner, mixed-alignment table. Flags as “CRITICAL UP / HIGH / LOW” and a separate critical-value notification footer. Date with time “18-May-2026 04:17”.
Normalized table
No reports ingested yet. Pick a sample above and click Ingest to watch the parser run and a normalized row land here.
The workflow, and how much of it is built here.
Solid steps run in this sandbox. Dashed steps are what a real implementation adds. Select any step to read what it does.
- 01Ingest
- 02Extract
- 03Normalize
- 04Route
Five lab reports
Built in this demoReports from five labs, each with its own layout, field names and date format. View the source of any of them before ingesting.
In detail
- →Ingests messy lab-report PDFs from 5 different labs — each with its own layout, field naming and date format.
- →Normalizes every report into a single operator-ready row: patient, age/sex, date, lab, study type, finding, ICD-10, flag.
- →Expands each row to show raw-extracted text side-by-side with the clean JSON record — so the parse is auditable, not a black box.
- →Surfaces an operator's note per case — what the lab parses is only useful if the routing rule around it gets actioned.
- →Each of the 5 reports keeps the layout quirks of the lab that issued it — a different patient line, a different date format, the diagnosis in a different place. Click View source to inspect the messy source.
- →Each sample ships with both the raw extracted blob and the normalized record — pre-mapped by claude-sonnet-4-5 and shipped as static JSON. No API at runtime.
- →The parsing console is a sequenced state machine that mirrors a real OCR → extract → normalize → flag pipeline; transitions are pure local state.
- →Fully client-side. No backend, no keys, no network calls — safe to ship as a static portfolio page.
- →Real OCR on scanned PDFs (Textract, GCV, or Tesseract) before Claude — most lab reports in the wild are flat raster scans, not text-embedded PDFs.
- →Per-lab extraction templates + a fallback Claude pass with structured-output schema, plus a confidence score per field for the human-review queue.
- →ICD-10 / SNOMED dictionary lookup with a curator step — the model proposes a code, a human accepts or overrides, and the pair becomes training data.
- →Critical-value routing — pager/Slack/SMS for flag = critical, with acknowledgement tracking and an SLA timer per case.
- →PHI handling: at-rest encryption, per-record audit log, retention policy, and a HIPAA/LFPDPPP review of the prompt + storage path.
- →Eval set per lab template — when a new layout drifts, the regex / prompt regression shows up before patients do.
The pattern transfers to any document-heavy operator workflow — invoices, lab reports, intake forms, contracts. The value is in the routing rules around the parse, not the parse alone.