Skip to content

One record at the centre, the assistants on top.

Your servers or ours.

Dudoxx is built in three layers around one patient record: the assistants a person talks to, the engines that do the work behind them, and the platform that keeps the record. Your data stays where you choose to run it.

A question arrives

A doctor asks Condor, the clinical engine, a question in the words used at the desk.

Before it answers, Condor shows its plan: the patient data it will gather and the steps it will take. The recap at the end lists what it did.

Asked at the desk

Differential diagnosis, please: 58-year-old man, pressing chest pain radiating into the left arm.

On the screen

  1. The question, exactly as it was asked
  2. The plan: phases and steps, shown above the answer
  3. Run recap: steps completed, tools used, time taken
A finished Condor run in a demo clinic: the question, the plan, the answer and the run recap.A finished Condor run in a demo clinic: the question, the plan, the answer and the run recap.
Real screenA finished Condor run in a demo clinic: the question, the plan, the answer and the run recap.

The engines behind them

Behind the assistants sit the engines that do the heavy work: language models, document reading, document generation, voice and video.

The language models are hosted by Dudoxx and reached through one router. There is no fallback to an outside AI provider.

  • Language models

    The reasoning behind each assistant, on models Dudoxx hosts itself.

  • Document reading

    Turns scanned letters and findings into text the record can search.

  • Document generation

    Drafts letters and reports from templates for a person to review and sign.

  • Voice

    Dictation and spoken requests, turned into text where the work happens.

  • Video

    Video consultations and meetings, with a virtual waiting room.

The gateway and the record

Every request passes one gateway first. It checks who is signed in, for which clinic and with which role, before any data is read.

The record is kept in FHIR R4, the open standard for health data, with one partition per clinic. Every change is written to an audit log.

  1. The desk

    A person, or an assistant working for them, asks for data.

  2. The gateway

    Checks the sign-in, the clinic and the role.

  3. The clinic's partition

    Only this clinic's data is within reach.

  4. The patient record

    Kept in FHIR R4; each change goes to the audit log.

Where it runs is your choice

Run Dudoxx in our cloud, on your own premises or as a hybrid of both. The screens your team works with are the same in each case.

Each clinic's data is isolated from every other clinic, and it is encrypted in transit and at rest.

How deployment works
  • Our cloud

    Start in the browser, with no server of your own to run.

  • Your premises

    On hardware the clinic owns, in a place your IT team decides.

  • Hybrid

    Our cloud and your premises combined, in the mix that fits the clinic.

The whole system in one diagram

The same layers as one picture, from the assistants at the top to the platform and the record at the base.

Dudoxx AI HMS architecture: one gateway, one clinical core, one patient record Clients (web portals, desktop app, mobile app) enter through a secure gateway into the clinical core (EHR, scheduling, documents, billing, messaging). Seven AI assistants work inside the core and use self-hosted language models. The core calls four processing engines and stores clinical data in the patient record, plus clinic data, files and a knowledge index. Live updates are pushed back to every screen. The platform runs as SaaS, on-premises or hybrid. AI ASSISTANTS · INSIDE THE PLATFORM Toucan first orchestrator Condor plans · asks · acts Falcon voice Parrot messages IBIS clinical guidance Phoenix help Sentinel paper to data plan · ask · act on the record PROCESSING ENGINES Private AI models self-hosted language models no outside AI provider Document reading scans to structured data Document generation branded PDF output Voice speech-to-text · text-to-speech Video visits consultations · meetings PLATFORM · FROM THE SCREEN TO THE RECORD Clients web portals · desktop app Secure gateway sign-in · roles clinic isolation Clinical core EHR · scheduling · documents billing · messaging Patient record one partition per clinic the single clinical source of truth Clinic data, files, index accounts · per-clinic storage search vectors YOUR INFRASTRUCTURE SaaS On-premises Hybrid live updates · to every screen patient record and EHR trust boundary data store request
Dudoxx AI HMS architecture: one gateway, one clinical core, one patient record Clients (web portals, desktop app, mobile app) enter through a secure gateway into the clinical core (EHR, scheduling, documents, billing, messaging). Seven AI assistants work inside the core and use self-hosted language models. The core calls four processing engines and stores clinical data in the patient record, plus clinic data, files and a knowledge index. Live updates are pushed back to every screen. The platform runs as SaaS, on-premises or hybrid. AI ASSISTANTS · INSIDE THE PLATFORM Toucan first orchestrator Condor plans · asks · acts Falcon voice Parrot messages IBIS clinical guidance Phoenix help Sentinel paper to data plan · ask · act on the record PROCESSING ENGINES Private AI models self-hosted language models no outside AI provider Document reading scans to structured data Document generation branded PDF output Voice speech-to-text · text-to-speech Video visits consultations · meetings PLATFORM · FROM THE SCREEN TO THE RECORD Clients web portals · desktop app Secure gateway sign-in · roles clinic isolation Clinical core EHR · scheduling · documents billing · messaging Patient record one partition per clinic the single clinical source of truth Clinic data, files, index accounts · per-clinic storage search vectors YOUR INFRASTRUCTURE SaaS On-premises Hybrid live updates · to every screen patient record and EHR trust boundary data store request
Dudoxx AI HMS architecture: one gateway, one clinical core, one patient record Clients (web portals, desktop app, mobile app) enter through a secure gateway into the clinical core (EHR, scheduling, documents, billing, messaging). Seven AI assistants work inside the core and use self-hosted language models. The core calls four processing engines and stores clinical data in the patient record, plus clinic data, files and a knowledge index. Live updates are pushed back to every screen. The platform runs as SaaS, on-premises or hybrid. AI ASSISTANTS · INSIDE THE PLATFORM Toucan first orchestrator Condor plans · asks · acts Falcon voice Parrot messages IBIS clinical guidance Phoenix help Sentinel paper to data plan · ask · act on the record PROCESSING ENGINES Private AI models self-hosted language models no outside AI provider Document reading scans to structured data Document generation branded PDF output Voice speech-to-text · text-to-speech Video visits consultations · meetings PLATFORM · FROM THE SCREEN TO THE RECORD Clients web portals · desktop app Secure gateway sign-in · roles clinic isolation Clinical core EHR · scheduling · documents billing · messaging Patient record one partition per clinic the single clinical source of truth Clinic data, files, index accounts · per-clinic storage search vectors YOUR INFRASTRUCTURE SaaS On-premises Hybrid live updates · to every screen patient record and EHR trust boundary data store request
Dudoxx AI HMS architecture: one gateway, one clinical core, one patient record Clients (web portals, desktop app, mobile app) enter through a secure gateway into the clinical core (EHR, scheduling, documents, billing, messaging). Seven AI assistants work inside the core and use self-hosted language models. The core calls four processing engines and stores clinical data in the patient record, plus clinic data, files and a knowledge index. Live updates are pushed back to every screen. The platform runs as SaaS, on-premises or hybrid. AI ASSISTANTS · INSIDE THE PLATFORM Toucan first orchestrator Condor plans · asks · acts Falcon voice Parrot messages IBIS clinical guidance Phoenix help Sentinel paper to data plan · ask · act on the record PROCESSING ENGINES Private AI models self-hosted language models no outside AI provider Document reading scans to structured data Document generation branded PDF output Voice speech-to-text · text-to-speech Video visits consultations · meetings PLATFORM · FROM THE SCREEN TO THE RECORD Clients web portals · desktop app Secure gateway sign-in · roles clinic isolation Clinical core EHR · scheduling · documents billing · messaging Patient record one partition per clinic the single clinical source of truth Clinic data, files, index accounts · per-clinic storage search vectors YOUR INFRASTRUCTURE SaaS On-premises Hybrid live updates · to every screen patient record and EHR trust boundary data store request

The Dudoxx product line in one diagram, from the assistants down to the platform.

Open the full diagram (opens in a new tab)

See the layers at work.

Book a demo and tell us which desk you want to see first.