LabMed in practice

Radiologists spend significant time writing reports and manually hunting for comparable prior cases; keyword search misses cases described with different terminology.

The system semantically ranks similar prior cases from the institution's own archive and assembles a structured report draft grounded in them. The physician reviews, corrects and signs the draft, and every corrected report immediately enriches the knowledge base. Everything runs locally; no data leaves the institution.

A structured draft in seconds instead of minutes of writing and searching; every signed report makes the next draft better; the archive stays in-house.

Patient therapy was tracked on a paper form and a ward Word document: no automatic safety checks, hard to read, and no audit trail.

The medication process was digitized from admission to discharge: a nurse photographs the handwritten form, AI handwriting recognition (with mandatory human confirmation of every field) captures the existing therapy, the physician reconciles and prescribes with a PIN e-signature, and administrations are recorded at the bedside, even offline.

Every dose, change and override is traceable to a person and a time; allergy checks run automatically; the ward keeps working when the network does not.

The monthly physician schedule across multiple sites and duty posts (ward, outpatient clinics, the catheterization lab, on-call shifts) was assembled by hand in spreadsheets: slow, error-prone, and with no guarantee that insurer slots and hour quotas were honoured.

An optimization solver generates an hour-by-hour schedule under hard rules (competencies, hour quotas, insurer windows, travel between cities) and layered priorities. When a slot stays uncovered, the system explains why and proposes concrete measures; for sick leave it automatically ranks substitutes with justification.

A month of schedules in minutes instead of days, with insurer windows and hour quotas guaranteed by construction, and a justified substitute proposed the moment someone calls in sick.

After a ransomware attack, a veterinary pathology institute was left with a damaged legacy database of more than 51,000 examination records, and no modern system for daily work.

The data was forensically recovered and turned into a secure, GDPR-compliant application: a searchable archive, structured report forms per examination type (necropsy, histopathology, cytology), image management, live voice dictation and AI report drafts, with inference running exclusively within the EU.

51,000+ records back in daily use and searchable in seconds; new reports dictated and drafted instead of typed; GDPR obligations automated rather than remembered.

A physician holding the result of a spectral biosignal measurement needs a fast, transparent ML assessment and a report they can share with the patient, without assembling it by hand.

The physician uploads the measurements, an ensemble of independent ML models votes on the classification in parallel, showing confidence and each model’s vote; generative AI composes the clinical narrative, and the PDF report lands in the patient's own secure portal.

The physician gets a classification with confidence and per-model votes, not a black-box score; the patient gets a readable report in their portal without anyone assembling it by hand.