Skip to content

Decision systems

Thousands of decisions go right. We build for the one that does not.

Avaxia makes software for work where being wrong is expensive. It reads the whole picture, shows where every number came from, and hands the decision to a person. We started in hospitals, because that is where the tolerance for error is lowest.

Talk to usSee what we build

The problem

The information was there. The system buried it.

Intensive care does not fail because clinicians lack knowledge. It fails because a person is asked to hold too much at once, and the software meant to help mostly shouts.

  • 23%

    of intensive care encounters involve an error in working out what is wrong. 17.8 percent of those caused direct harm.

    Auerbach et al., 2024

  • 64.1

    passive alerts per intensive care patient, per day. Only 4.5 percent are genuine emergencies.

    J. Patient Safety

  • 96%

    of computerised medication alerts are overridden by clinicians, at the high end of published studies.

    PMC systematic review

  • 1.7

    medical errors a day for the average critically ill patient, some of them life threatening.

    medRxiv, 2025
Every hour of delay in antibiotics after septic shock costs 4 to 6 percent of estimated survival.Ku et al., 2025
100857055100%62%+0h+1h+2h+3h+4h+5h+6h
Every hour of delay in antibiotics after septic shock costs 4 to 6 percent of estimated survival.
0 hours after onsetEstimated survival, percent
0100
192
285
379
473
567
662

On these numbersEvery figure above is published research about the problem, with its source named. None of them is a claim about what Avaxia’s software achieves.

What we build today

Two products. Both show their working. Neither one acts.

Both are in healthcare, and both are about medicine and the decisions around it. The approach is not specific to medicine, and the next product may not be.

Accu-RxInside the record system, and standalone

Every prescription checked before it is filled.

Drug-to-drug interactions and drug-allergy conflicts, caught at the moment of prescribing. It runs inside the hospital record system for hospitals and insurers, and stands alone for clinics and people who do not have one.

$42B
estimated global annual cost of medication errors.WHO, 2024
1.3M
patients a year experiencing harm related to their medicines.WHO, 2024
HOLDAccu-Rx · order check · 03:14

Being ordered

Ciprofloxacin 400mg IV

q12h, 7 days

Already active

Tizanidine 4mg PO

started 2 days ago

CYP1A2 inhibition raises tizanidine exposure several-fold. Expect a marked drop in blood pressure and sedation. The prescriber confirms or picks an alternative.

What Accu-Rx does

CritivaIntensive care

A whole plan for an ICU patient, with its sources attached.

Critiva reads vitals, labs, medications, imaging and notes, then proposes a full assessment and plan against current international guidance and the hospital’s own history. Anything conflicting with the record is withheld before a doctor sees it.

$878B
estimated global annual cost of unsafe care, 12.6 percent of health spend in high-income countries.WHO, 2024
~35,600
US intensive care patients a year estimated to die with a finding that was missed.Auerbach et al., 2024
Critiva · assessment & planCase 007 · day 3

  • Proposed

    Noradrenaline to 0.18 mcg/kg/min

    MAP 58 across 3 readings · 74kg, 12 Aug

  • Proposed

    Hold enoxaparin, recheck platelets 06:00

    Platelets 71 to 48 in 24h · labs 03:02

  • Withheld

    Vancomycin 1g q12h, removed before review

    Documented anaphylaxis · admission note

  • Proposed

    Cardiology review within 12 hours

    Troponin rising, new wall motion abnormality · echo 21:40

Critiva at critiva.health

How we build

Four rules, each enforced somewhere a person cannot quietly skip it.

None of these are specific to medicine. They are what we would carry into any field where a mistake is costly. Each one names the place it is enforced, because a rule that lives only in a document is a preference.

  1. Standards first, never guesswork

    Every dose, score and risk calculation runs against current international guidance, in code. No arithmetic is left to a language model.

    Enforced inIn code

  2. The software proposes, a person authors

    Nothing is acted on without a person putting their name to it. That is a boundary in the architecture, not a setting somebody can change.

    Enforced inIn the architecture

  3. Guardrails are code, not instructions

    A rule written into a prompt is a request. A rule written into the pipeline is a rule. Ours are in the pipeline, and they are tested.

    Enforced inIn the pipeline

  4. Measured against an expert’s own answers

    Output is scored against reference plans written by a practising intensivist, on cases held back from development.

    Enforced inIn the evaluation harness

Where this goes

Healthcare is where we started, not the boundary.

The hard part was never the medicine. It was building software that reads a messy, incomplete record, works out what it can prove, refuses to guess the rest, and leaves the decision with the person accountable for it. That problem exists in plenty of other rooms.

For now it has taken us into a category that is growing on its own, and into three markets we already understand.

$5.8B
global clinical decision support market in 2026.MarketsandMarkets, June 2026
$10.15B
the same market in 2031, growing 11.8 percent a year.MarketsandMarkets, June 2026
$7.3B
estimated annual US cost of preventable adverse drug reactions.WHO, 2024

  • Palestine

    First mover, plugging into Avicenna hospital systems directly.

  • Jordan

    Standards-based integration alongside Hakeem, without replacing it.

  • International

    Built for interoperability from the start, rather than a single-market bet.

Who is behind it

Built in Hebron, by people who have shipped software and practised medicine.

  • Ayman Irziqat

    Co-founder & CEO

    Doctorate in digital transformation, and 25 years in leadership and sales.

  • Nazeeh Irziqat

    Co-founder & CTO

    Senior software engineer in artificial intelligence and data analytics, working on healthcare applications.

  • Abdullah Zreqat

    Co-founder & software engineer lead

    Builds the engine, the safety gate and the evaluation harness behind Critiva.

  • Dr. Raed

    Clinical validation lead

    Practising intensivist. Reviews every protocol and guideline the engine recommends at the bedside.

Talk to us

We would rather show you than tell you.

We will walk you through what the software returns on a real case, including the parts it gets wrong. If you run a unit, a pharmacy, an insurer, or a fund, start here.

Start a conversationRequest the technical brief