The infrastructure for clinical evidence.

Latrikos turns unstructured clinical information into longitudinal, multicentre and auditable databases, ready for Real-World Evidence, clinical registries and clinical trials.

Clinical AI in real time

It all starts with a patient.

A visit, an examination, a report. Clinical practice already produces the data research needs every day — and Latrikos asks it to change nothing.

And ends in free text.

The doctor writes what they see: diagnoses, treatments, progress. All of it correct, all of it useful — and all of it trapped in prose no database can read.

That is where evidence stops.

Building a study means re-reading everything and keying it into Excel or RedCap by hand. Months of work, projects delayed and variables lost along the way.

This is where Latrikos begins.

Drop the reports in exactly as they are. No templates, no integrations and no change to the hospital's workflow.

Text becomes data.

Every sentence is read in its clinical context and turned into variables traceable to the exact line they came from, with 99% accuracy.

From months to minutes.

Ready cohorts, comparable variables and results that compute themselves: ten times faster and at 80% less cost.

And no hospital researches alone.

Multicentre studies on a shared standard, where every site contributes its data without giving up control. Real-world evidence, finally at scale.

ComplianceLatrikos
ENSSpanish National Security Framework
GDPRData protection
HIPAAHealth security and privacy
OSSIInformation Systems Security Office
InfrastructureCertified
ISO/IEC 27001Information security management
ISO/IEC 27017Cloud services security
ISO/IEC 27018Personal data in the cloud
ISO/IEC 27701Privacy information management
ISO 50001Energy management
ISO 14001Environmental management
ISO 9001Quality management
HDSHealth data hosting
The platform

The medical record is finally a database.

Every report written in your hospital holds valuable variables trapped in free text. Latrikos frees them in real time and keeps them ordered, traceable and ready to query — without changing clinical practice.

  • Built for GDPR
  • On-premise or cloud
  • End-to-end encryption
  • Interoperable · HL7 FHIR
The problem

The clinical data exists. It just cannot be used.

Four bottlenecks that repeat across every department, and that no record template has ever solved.

Translucent clinical documents piled up in disarray, with valuable data trapped inside.
01

Free text, locked in

The hospital’s richest information lives in paragraphs no SQL query can read.

02

Hours of manual entry

Structuring records by hand consumes the time of clinicians, data managers and residents.

03

Incomplete records

What gets transcribed is a biased fraction; the rest never reaches any analysis.

04

Research held back

Every study starts months late because the dataset has to be built from scratch.

Product

Three modules, one flow.

From the raw document to the published study, with no manual exports in between.

The Latrikos flow: from raw documents to structured data and a connected living database.
01

Extraction

From reports, notes or test results, the AI identifies every clinical variable instantly, with its context and its source.

02

Living database

Variables settle into a consistent schema that grows with every document and audits back to the original text.

03

Research

Define cohorts, launch multicentre studies and follow the evidence in real time, with no manual exports.

In practice

Your department, measured live.

A single panel showing processed volume, variable coverage and the status of every active cohort.

latrikos · department panel

7-week rolling window

Documents processed per week

01k2k3k4kS-07S-06S-05S-04S-03S-02S-013 240

Variable coverage

Primary diagnosis99.4%
Active treatment98.7%
Medical history97.9%
Pending review0.6%
Coverage95% target threshold
Results

What changes when data flows.

Measured across departments that have already structured their full history with Latrikos.

Study speed10x

faster to assemble a research dataset

Extraction accuracy99%

of clinical variables correctly structured

Time recovered+80%

of manual entry time returned to the clinical team

Benchmark

Accuracy and speed, without choosing.

Evaluated on a real clinical corpus against manual extraction, rule-based systems and generic NLP.

Accuracy versus processing speed

n = 4 approaches · bars = ±1 SD
Extraction accuracy · %Processing speed · documents/hour →6070809010003006009001 2001 500Manual extraction96.5% · 14 doc/hRule-based systems71.0% · 430 doc/hGeneric NLP83.5% · 820 doc/hLatrikos99.0% · 1 240 doc/h
Evaluated on a real clinical corpus. Bars represent ±1 standard deviation across batches.
Trustworthy AI

Nine principles, met by design.

Latrikos is built on the European principles for trustworthy AI. They are not a statement of intent: each one maps to a concrete technical decision.

Human agency and oversight

Human-in-the-loop: the researcher co-designs the variables, validates before and after, and checks every result against the anonymised source text.

Technical robustness and safety

Continuous validation, SAST/DAST audits and ENS compliance. Access control, traceability, backups and incident management.

ENS · Basic level

Privacy and data governance

De-identification and pseudonymisation up front. Original documents are deleted after processing: only structured variables persist.

Privacy by design

Transparency

Every variable keeps a reference to the anonymised fragment it came from. Full traceability and a verifiable result.

Diversity, non-discrimination and fairness

The tool does not classify or decide about people. It only structures clinical information that already exists.

Societal and environmental well-being

Secondary reuse of clinical data on energy-efficient infrastructure.

Energy management

Accountability

Documented governance, defined roles and auditing of access and operations across the full processing cycle.

Human dignity

Proactive training for the research team. AI amplifies their work, it does not replace it.

Magnifica Humanitas

AI Act

Latrikos neither makes nor intervenes in clinical decisions: it structures existing information for research purposes.

Low-risk category

Let’s design the path from your clinical histories to evidence.

We only ask for what we need to prepare the first conversation. We design the project together — you do not have to bring it solved.

Turn your next report into your next study.