Clinically fine-tuned model
Diagnify selects licensed open-weight base models and clinically fine-tunes them on a curated corpus derived from historical consultations, with correction and feedback from clinical reviewers.
Diagnify clinician product · in evaluation
AI clinical decision support that keeps the clinician in control.
Diagnify is the clinician-facing product built on a selected, licensed open-weight model clinically fine-tuned by Diagnify, with relevant guidance retrieved separately from a maintained evidence database. It helps qualified health professionals organise a case, check red flags, compare differential diagnoses and plan investigations; it does not autonomously make diagnoses, prescriptions or clinical decisions.
Medical intelligence foundation
One case. Two intelligence layers. One accountable reviewer.
Historical clinical material shapes the fine-tuned model. Current guidelines and source-linked evidence are retrieved from a separate database. The clinician reviews both in the context of the patient.
Guideline and evidence retrieval
At inference, relevant material is retrieved from a separately maintained evidence database rather than relying only on information learned during fine-tuning.
Qualified clinician review
The resulting assessment is made reviewable so the clinician can test its assumptions, evidence, uncertainty and fit for the patient.
Corpus description: Diagnify currently describes its internal clinical fine-tuning corpus as being derived from 50M+ historical consultations spanning 50 years. This company-supplied description refers to the training corpus, not the separate guideline and evidence database. Read the methodology and limitations.
Product overview
A structured second set of eyes for the clinical work-up.
Diagnify is intended to support—not replace—the clinician’s reasoning. Its workflow applies Diagnify’s clinically fine-tuned model and separate evidence-retrieval layer to the questions that matter during a consultation, rather than sending a prompt to an unadapted general-purpose chatbot.
Structure the presentation
Bring symptoms, history, examination findings and clinical context into a consistent problem representation.
Challenge the differential
Consider likely, important and easily missed possibilities while keeping uncertainty visible.
Verify before action
Review red flags, evidence, assumptions and local guidance before accepting or acting on an output.
Clinical workflow
From presentation to a reviewable reasoning record.
-
Enter the relevant clinical context
Use the minimum information needed for the task and follow your organisation’s privacy and consent requirements.
-
Review the structured assessment
Inspect symptom framing, risk features, candidate explanations and the assumptions that influence prioritisation.
-
Interrogate the evidence
Check source relevance, recency, population fit and uncertainty. A citation is a starting point for verification, not proof that a recommendation applies.
-
Apply professional judgement
Confirm conclusions against the patient, local guidance and scope of practice. Reject or revise any output that does not fit.
Evidence visibility
Designed for inspection, not blind acceptance.
Evidence-based clinical support requires more than attaching references to fluent text. At inference, Diagnify retrieves relevant material from a separately maintained evidence database so clinicians can inspect the basis for an output and recognise where evidence is incomplete or indirect.
- Evidence links presented in context
- Separation of case facts from model inference
- Visible uncertainty and important alternatives
- Prompts to check recency, applicability and local guidance
Private platform access
Building a medical-AI product? Build on a dedicated deployment.
Approved medical-AI teams can request a separately hosted deployment of the clinically fine-tuned model and evidence service for use in their own products and workflows.
Model plus evidence service
Private access combines the clinical model with the separately maintained guideline and evidence-retrieval layer, delivered for an approved and defined medical use case.
Customer isolation
Customer prompts, outputs, uploaded data and customer-specific configurations are isolated from other customer deployments. They are not used to train or serve another customer’s deployment without explicit written authorisation.
Intended professional use
A reasoning aid for qualified health professionals.
Appropriate use during evaluation
- Supporting a clinician-led assessment
- Testing the completeness of a differential
- Identifying questions or findings to verify
- Reviewing evidence alongside authoritative guidance
Not an appropriate use
- Autonomous diagnosis, triage or treatment
- Emergency decision-making
- Direct-to-patient medical advice
- Replacing examination, local policy or specialist review
Evaluation status: access to a product in testing does not establish clinical effectiveness, fitness for a particular purpose or regulatory authorisation. Organisations should perform their own governance, validation and risk assessment.
Explore the system
Evaluate Diagnify from workflow to evidence.
Clinician access
Evaluate the workflow with its limits visible.
Use Diagnify only within your professional scope and under appropriate organisational governance.