Skip to content
Lia
Industries

AI that a clinician will actually rely on.

iLeaf has built healthcare software since 2011, from BMJ's digital platform to diagnostic tools and a clinical AI stack that cut specialist misrouting by 59%. In healthcare the constraint is not capability but consequence, so we run on-premise where required — one deployment kept every byte of patient data inside the hospital.

Talk to a solutions lead See client results
Ask Lia about healthcare

What this includes

Clinical AI
Intake triage, specialist matching and record summarisation, always with the clinician making the final call.
EMR & HL7/FHIR integration
Working with the record systems you already have, including the older interfaces that are not going anywhere.
On-premise deployment
The entire stack, models included, running inside your infrastructure when data cannot leave the building.
Patient-facing products
Apps patients actually use — including diagnostic imaging tools and a jaundice screening application.
Publishing & evidence platforms
Large clinical content and evidence systems, the work behind our engagement with BMJ.
Audit & explainability
Every AI-assisted decision reconstructable afterwards, because in clinical settings that is not optional.

How we run it

Every phase ships something usable on its own, so you are never holding a half-finished system waiting on the next milestone.

  1. Start where being wrong is survivable

    First deployments go to routing, summarisation and admin load — not diagnosis. Trust is earned on lower stakes.

  2. Keep the clinician deciding

    The system proposes and evidences; a person decides. Nothing autonomous touches clinical judgement.

  3. Design for the data boundary

    Where the data may live is settled before architecture, not retrofitted after a security review.

  4. Measure clinical outcomes

    Success is misrouting, waiting and consultation time — not model accuracy in isolation.

What we build it with

  • HL7 / FHIR
  • On-premise GPU
  • Open-weight models
  • Python
  • PostgreSQL
  • Graph-RAG
  • Swift
  • Kotlin
  • DICOM
  • Docker
  • Governor Engine
  • OpenTelemetry

Questions we get asked

Can AI run without patient data leaving our network?

Yes, and we have done it. A full agentic stack — retrieval, models and governance — ran entirely on a hospital network's own GPUs, with no patient data transmitted to any external service. Open-weight models make this practical without giving up meaningful capability.

How do you handle regulatory expectations?

By designing for the audit from the start: every AI-assisted step records its inputs, reasoning and outcome, and clinicians remain the decision-makers. We work to the frameworks your compliance team already operates under rather than proposing a parallel process.

Do you have real healthcare delivery experience?

Since 2011. That includes BMJ's digital transformation, a jaundice screening application, diagnostic imaging work with Picterus — which raised over $3M from the European Innovation Council — and the clinical AI platform quoted above.

Related

Let’s talk about healthcare.

Tell us what you are running and what it needs to do next. We will tell you honestly whether we are the right team for it.

Talk to a solutions lead
Selected work

Healthcare: the work behind this