Document intelligence for regulated industries: lab PDFs, inspection evidence, funder reports. We extract what is on the page — and refuse to guess when it is not.
Geotechnical consultancies still re-key Certificates of Analysis. Registered managers still assemble CQC evidence in shared drives the week before an inspection. CICs still rebuild funder tables from spreadsheets that do not match last quarter’s data import sheet. Oil and gas handover packs still arrive as mixed PDFs and vendor data books. None of that is an MVP pitch-deck problem. It is a document + evidence problem.
minidev.pro builds the software layer on top of those documents: automated CoA screening, CQC compliance readiness tools, lab PDF data extraction, funder impact reporting software, and evidence packs for Ofsted safeguarding, DICOM imaging governance, and WCAG conformance. Same senior engineers who ship 30-day MVPs — different domain, higher cost of being wrong.
Automated CoA screening against GAC, S4ULs and C4SLs. Determinands extracted from UK lab PDFs — Chemtest, i2, DETS, SOCOTEC and the rest — without inventing a mg/kg value.
Self-assessment against quality statements and the five key questions. Predicted ratings with explanations, action plans, role-based access for registered managers and staff.
Lab PDF data extraction, oil and gas handover documentation, DICOM and medical imaging governance, WCAG evidence packs. Parsers that fail closed.
Participant pathways from referral to employment, outcome recording with evidence, reports that match the funder’s data import sheet — not a generic dashboard.
Four UK lab PDF formats, column clustering from pdf.js word positions, 148/148 gold cells, dual-parser cross-check.
32 quality statements, five key questions, predicted ratings, action plans, RBAC for registered managers.
CIC pathways, evidence-backed outcomes, funder-ready report generation.
Policy engine, approval routing, spend and compliance reporting.
We start from the artefacts you already produce or receive: CoAs, quality-statement evidence, KLOE-era folders people still use as a mental model, data import sheets, vendor data books. Then we lock a schema — determinands, units, sample IDs, quality statements, pathway stages — before any UI. Extraction and scoring come next. Humans review exceptions. The system does not fill gaps with a model’s best guess.
Typical first slice: one document family, one scoring or reporting output, role-based access, audit log. 30 days if the corpus is bounded. After that, retainer work adds labs, sites, or funders.
Consultancies, providers, and CICs who already know their regulations and are tired of re-keying. Not a fit if you want a chatbot that “summarises compliance.” Read the case studies before a call.
We will tell you whether they are parseable, and what we would not automate.
→ Book a call