About

We build testable software for medication safety and billing.

LavieuxLabs is an R&D team bringing together clinical pharmacology, healthcare operations and software engineering. For work where a mistake is costly to a patient or an institution, we build systems whose every result can be reproduced and that leave the decision to a person.

Vision

The reason behind every clinical and administrative decision should be visible.

A decision support system is worth what it can reproduce and explain, not how clever it looks.

Mission

Testable software where errors are expensive.

In medication safety and billing, we build tools that make the physician's and specialist's work easier, not tools that replace them.

Why rule-based

At the point of decision, rules rather than predictions.

Probabilistic and generative models are useful for research and analysis. But an alert that can stop a prescription has to give the same result every time and show where it comes from. That is why we use a rule-based system at the point of decision.

Reproducibility
ProbabilisticThe same input can produce different outputs.
Rule-basedThe same input and rule version always produce the same output.
Explainability
ProbabilisticThe reason may be an explanation generated after the fact.
Rule-basedThe reason is the rule that fired.
Verification
ProbabilisticBehaviour is estimated statistically, from a sample.
Rule-basedEach rule is verified directly against defined test cases.
Change control
ProbabilisticRetraining can cause unexpected changes in behaviour.
Rule-basedEach change is versioned, reviewed and regression-tested.
Failure mode
ProbabilisticFluent but wrong output (hallucination) can be hard to spot.
Rule-basedCases outside scope are marked plainly as “not assessed”.

How we work

From source to code, from code to test.

We apply the traceability and change control a regulated product needs from the first day.

Rules with a known source

Rules are taken from official product information (SmPC and patient leaflet), clinical guidelines and peer-reviewed literature. Every rule shows its source.

Sources
SmPC / PIL · guidelines · peer-reviewed literature
Traceability
Rule → source

Traced from requirement to test

Each clinical requirement is linked to the rule that implements it and the test that checks it.

Chain
Requirement → rule → test
Automated tests
2,480+

Controlled change

Rule sets are versioned. No change is released without review and passing tests.

Versioning
Rule set · knowledge base
Gate
Review + regression tests

Security and privacy from the start

Collecting little data, role-based access and an unchangeable log were not added later; the system started with them.

Controls
Data minimisation · RBAC
Log
Append-only audit trail

Quality and regulation

Which standards, and where we stand.

We list the standards we work against and where we stand on each. This is not a certification statement.

REG-01Class IIa target

EU Medical Device Regulation

PharmaDeux provides information used in drug therapy decisions, so it falls under the MDR.

Regulation
MDR 2017/745 · Annex VIII Rule 11
Target
SaMD · Class IIa
CE mark
Not yet
REG-02Architecture aligned

Software life cycle and risk

Development and risk management processes are being set up to these standards.

Life cycle
IEC 62304
Risk
ISO 14971
REG-03On the roadmap

Quality management system

The quality management system and usability work are on the roadmap.

QMS
ISO 13485
Usability
IEC 62366-1
REG-04Control set target

Information security and personal data

Personal data is processed under KVKK (Turkey's data protection law); security controls are being set up to ISO/IEC 27001.

Law
KVKK No. 6698 · Art. 12
Controls
ISO/IEC 27001

Clinical collaboration

We build it with the teams who will use it.

A decision support system cannot be safe if it does not fit the workflow of the team using it.

  • Clinical advice

    Rule sets and screens are written and reviewed with clinicians and clinical pharmacists.

    Stakeholders
    Clinician · clinical pharmacist
    Output
    Reviewed rule sets

  • Ethics committee and data

    Every study on real data needs ethics committee approval, anonymisation and a data processing agreement with the institution.

    Prerequisite
    Ethics approval · DPA
    Data
    Anonymised, KVKK-compliant

  • Publication

    We aim to publish validation studies in peer-reviewed journals together with partner institutions.

    Goal
    Peer-reviewed publication
    Reporting
    Sensitivity · specificity · PPV

  • Feedback from the field

    Feedback from pilot users is collected regularly, and we track which requirement it turns into.

    Collection
    Structured form
    Traceability
    Feedback → requirement

Contact

Write to us about a pilot, a validation study or academic work.

If you have an idea for a clinical, institutional or academic study, a short message is enough.

Write to us