Fifty-four working demonstrations of document extraction, structured transformation and validation workflows — see the full portfolio overview.
Fully documented end to end — inputs, pipeline, validation discipline, committed test evidence, and known limitations. See the full detail for each. The rest of the catalog below is supporting detail.
Tags free-text items against an operator-defined taxonomy, then proposes and independently re-verifies links between related items.
Converts a real .docx manuscript into real NISO JATS 1.3 XML, validated against the official schema from NLM/NCBI.
Deterministic structure detection (headings, references, captions) from a real PDF's own extracted text, with an AI-assisted per-section description.
Anchor-verified field extraction against a real lease-specific schema, plus real lease-domain checks (date order, escalation, deposit ratio).
Fixed deterministic detectors plus operator-defined custom criteria, with every AI-flagged quote independently verified against the real content.
Supporting detail beyond the five above.
AI-assisted proofreading with human-in-the-loop review.
Schema-driven extraction from a list of URLs, with AI confidence scoring and human review.
PDF accessibility tag-structure checking (not just visual appearance), with AI-assisted heading detection and human review.
Invoice arithmetic verified in code, plus PO/vendor cross-checks, with human review.
Real OCR (Tesseract) with per-word confidence scoring and human-in-the-loop correction.
DOCX to real EPUB 3.0, QC'd against independently-read source stats, with human review.
AI-assisted text classification against a custom label taxonomy, with human-in-the-loop review.
AI-parsed scheduling requests verified against real calendar logic, with human-in-the-loop review.
Presentation title/language, slide title placeholders, image alt text, and tables, with AI-assisted visual-title detection and human review.
Document title/language, heading outline, image alt text, and table header rows, with AI-assisted heading detection and human review.
Real local speech-to-text (Whisper) with timing, caption-format QC against broadcast guidelines, and human-in-the-loop review — exports real .srt/.vtt.
Real silence detection finds narration gaps; Claude proposes a description from the actual frame; real text-to-speech measures whether it fits, with human-in-the-loop review.
Fixed automated detectors (SSN/email/phone, a Luhn-validated card-number check, profanity) plus operator-defined custom criteria Claude screens for — every AI-flagged quote is verified against the real content before it's shown, with human-in-the-loop review.
Claude proposes a label per image from an operator-defined taxonomy; a reviewer confirms or corrects before export — same human-in-the-loop pattern as Text Annotation, applied to images.
Claude drafts WCAG 1.1.1 alt text per image; a deterministic QC layer flags empty text, redundant "image of" phrasing, over-length drafts, and filename-as-text before a reviewer edits and confirms.
No AI call — real local TTS renders each paragraph to real audio; a real pacing check compares every section's measured seconds-per-word against the job's own median to catch synthesis failures, with human-in-the-loop review.
Real file metadata (dimensions, resolution, color space) read from the artwork itself and checked against an operator-defined print spec — no AI step, every finding a plain comparison.
A real LaTeX tokenizer separates prose from markup — comments, math, and code are never sent to Claude — so only the actual text gets proofread, never command syntax or citation keys.
Real term-frequency analysis finds repeated named concepts with their real page numbers; Claude suggests conceptual topics a frequency count can't catch, each one verified against the real page it claims to be on.
A regex parser built from a known set of U.S. case-law reporter abbreviations extracts and reformats legal citations; every normalization is re-parsed and compared component-by-component against the original.
Claude structures a case into standard brief sections (Facts, Issue, Holding, Reasoning); every point is required to cite a quote checked against the real source — an unverifiable claim is flagged, not hidden.
Same shape as Invoice Verification applied to insurance claims — real arithmetic, date-range, and policy-limit checks, with the real payable amount computed in plain code.
Two real conversion pairs: EPUB2 to accessible EPUB3 (real nav.xhtml and accessibility metadata, built from the source's own NCX) and Word to DocBook 5 XML (real nested sections built from actual heading levels) — every output re-parsed and independently verified against the source, no AI involved.
Reads a real SCORM package's own manifest to find each real content page, then checks language, title, image alt text, and heading structure on each one — plus manifest integrity checks for resources the course references but never actually ships. No AI involved.
Real multi-speaker text-to-speech (espeak-ng, no AI call) in genuinely distinct voices per speaker, stitched into one real MP3 with real embedded chapter markers — independently re-read with ffprobe afterward and checked against what was actually declared.
No AI step — each billed shipment is looked up against a real contracted rate schedule (carrier, service level, zone, weight tier, real effective date window) and the real charged amount is compared against the real contracted rate, catching overcharges, expired rates, and duplicate billing.
A real word-level diff finds every real edit between two document versions — a reprint against the original, or one edition against the next. Claude classifies each edit, and the classification itself is checked: a count mismatch, an uncategorized-edit check, and an independent word-count reconciliation against the two full real texts, kept honest against the demo's own per-job classification cap.
No AI in the structure pass — a real deterministic sweep finds a paper's section headings, reference list, and figure/table captions from the file's own real extracted text. Claude adds a one-sentence description per real section, matched verbatim back to the real heading list; a paraphrase, wrong page, or invented heading is dropped and counted, never shown as if it were real.
A single real whole-document score, not a per-field confidence number. Combines real exact-phrase checklist coverage, real Flesch Reading Ease readability, and a consistency score that penalizes real, anchor-verified contradictions Claude finds and quotes — a fabricated quote is dropped and counted, never shown. The combination itself is independently recomputed from the three stored sub-scores before being trusted.
Real transformative image editing, not metadata reading. Claude (vision) proposes regions to redact; every box's real geometry is validated against the real image bounds before a reviewer sees it. Only confirmed regions get a real solid-black rectangle composited onto the real pixel data, then independently verified by measuring the real output pixels directly and sampling outside every confirmed region to catch bleed.
Case Summarization's anchor-verified extraction pattern applied to a real lease-specific field schema, plus real lease-domain checks a generic extractor doesn't have: a real lease-term end date after the start, a Rent Escalation schedule checked for a real decrease, a Security Deposit checked against Base Rent for a plausible ratio, and coverage of the fields every real lease states.
No AI for the two criteria this line names specifically: a real difference hash of each uploaded image's own pixel data, compared against every other image in every real orientation, catches a reused or duplicated figure; two real exact arithmetic checks recompute a stated count/percentage triple and check a percentage breakdown against 100%. Claude adds disclosure-statement screening corroborated by an independent keyword search, and numeric-contradiction detection anchor-verified on both quotes.
Text Annotation's proven schema-constrained tagging, plus the real "linking" half it never built: Claude proposes pairs of genuinely related items, each required to quote a real supporting phrase from both sides. Every link is independently re-verified, with a link to a nonexistent item, a self-link, an unverified anchor, or a duplicate pair dropped and counted before a reviewer sees it — and a real, AI-free lexical-overlap score shown alongside every link, never used to reject a genuinely related pair.
A real PRISMA-style inclusion/exclusion protocol, not a single taxonomy label. Claude reports Include, Exclude, or Maybe per record; an Exclude is only valid alongside one real reason from a controlled list, an Include only alongside one real inclusion criterion — either independently downgraded to "maybe" when the pick isn't actually one of the real listed options. The PRISMA flow diagram is computed live from the records' own current decisions, never cached, with a reconciliation check confirming the exclude count equals the sum of its per-reason breakdown.
Verified against a real SEC EDGAR filing, not an invented one. A no-AI section-locator finds the real litigation-disclosure heading in a live-fetched document; Claude reports every distinct patent-litigation matter (a district-court suit, an ITC Section 337 investigation, a PTAB proceeding); a real, closed list of all 94 federal judicial districts plus ITC/PTAB/Federal Circuit independently validates every forum; every patent number is checked for valid format and separately cross-checked against its own record's quoted anchor text, catching a claim anchor-verification alone would miss.
Converts a real .docx manuscript into the real NISO JATS 1.3 Archiving and Interchange Tag Set, a different, specific XML vocabulary from DocBook used by PubMed Central and open-access journals. The generated XML is validated against the real, official JATS 1.3 XSD schema from NLM/NCBI, checked via a real libxml2 build (xmllint-wasm), not a hand-built approximation. Claude's one job is classifying each reference's real publication type, independently re-checked against the real JATS Tag Library's controlled vocabulary and the citation's own anchor text.
Fetches a real U.S. court opinion live from govinfo.gov's official "United States Courts Opinions" collection via its real public REST API — real structured metadata (court, parties with real roles, case number) straight from the government's own fields. Claude's one job is identifying the real disposition from the real opinion PDF text, independently re-verified against a real closed vocabulary and the opinion's own anchor text, with every party role in the real API response separately checked against a real, closed list of standard U.S. court-party roles.
Claude writes a generic abstract and keyword list for any document, not a structured legal brief. Every keyword is independently re-verified as a real, verbatim substring of the source text, with a case/whitespace-variant repeat caught as a duplicate, and the abstract is separately checked for a real compression-ratio bound and for any standalone number it states that doesn't actually appear in the source.
Claude translates a pasted English abstract into a real target language, then translates that output back into English without ever seeing the original — a blind round trip so the check is genuine, not an echo. Every number in the original is checked against the back-translation, with a real length-ratio bound and a lexical-overlap transparency signal rounding out the check.
No AI step — every real page break comes from actually rendering the uploaded .docx through LibreOffice Writer, the same real engine PowerPoint/Word Accessibility Audits already use. Every real Word heading is matched to the real page it first appears on, and the same verified page numbers are applied to a paginated XML skeleton too.
No AI step — checks a PDF's own embedded image objects, extracted directly via poppler's pdfimages, or a directly-uploaded image set, for real duplicates using a real perceptual difference hash in every real orientation, reusing Research Integrity Support's own proven hashing engine generalized beyond its single-manuscript scope.
No AI step — every real composition property is read directly from the PDF with poppler (real page geometry, real per-font embedding status, real per-image effective print resolution and color mode) and compared against an operator-defined spec, reusing Artwork Processing's own real-metadata-vs-spec discipline.
No AI step — every real paragraph's real applied font and size, and the document's own real page margins, are read directly from the RTF's own control words with a real pure-JS parser (no system-binary dependency) and compared against an operator-defined spec.
Not a checker — the first app in this portfolio to actually apply a real template (page size, margins, default body font) directly to a .docx's own OOXML and write a new file, independently re-verified three separate ways including a real LibreOffice open/convert attempt.
No AI step — a real affine transform fit by least squares from operator-given control points, independently re-checked via real meter-converted residuals, exported as a real standard GIS world file, and used to convert real points of interest into real GeoJSON.
No AI step — every real EPUB property (package metadata, manifest, spine, navigation TOC coverage, per-content-document structure, real embedded-image dimensions and format) is read directly from the archive's own XML and checked against an operator-defined spec, the same real-metadata-vs-spec discipline PDF Composition and QC and RTF Styling already use.
Every real PDF page is rasterized and sent to Claude for sensitive-region detection; confirmed regions get a real pixel composite, and the whole document is reassembled as a flat image-only PDF with zero extractable text, independently confirmed via poppler's own pdftotext.
No AI step — real workbook title/language, per-sheet tab names, real Excel Table coverage, merged-cell ranges, and per-image alt text read directly from the archive's own OOXML and checked against WCAG-cited rules, the same structural-audit discipline the Word and PowerPoint accessibility audits already use.
No AI step — a real .docx converted to a real PDF with LibreOffice Writer, independently checked by three separate tools: real source vs. output word counts, real page count, and real per-font embedding via poppler.
No AI step — real PDF AcroForm fields read, updated, and independently re-verified with pdf-lib. An update naming a field that doesn't exist is caught, and an update selecting a dropdown option outside its own declared option list is rejected by hand, since pdf-lib itself accepts it silently.
No AI step — real XPath-targeted updates against the document's own structure, each immediately re-validated against its own real XSD schema and rolled back on the spot if it would break validity. An XPath matching zero or more than one real node is rejected rather than guessed at.
Claude organizes free-text content into an operator-defined, multi-column table, every value required to carry a real verbatim quote from the source — independently checked against the real source before a reviewer sees it, with an unverified quote, an empty value, or a column outside the defined structure caught and flagged.
Claude proposes a real pixel position, size, and font size for every content block; plain geometry independently checks the real page margin, every pair of blocks for a genuine overlap, every font size against its own defined range, and the real reading order and gap — a real bug was caught live when the API returned one block with an incomplete geometry object.
No AI step — real rows grouped by a real composite key into exact and conflicting duplicates, one representative row per key compared pairwise with a real Levenshtein similarity score for near-duplicate typos, and a migration-ready CSV export that never silently drops data, only marking a row once a reviewer has actually confirmed the finding.