Thalix Health

Video · 3:08 · synthetic data

Thalix walkthrough

A synthetic demo hospital end to end: feeds in one replayable log, the same patients seen differently by role, an assistant that proposes and a person who confirms, a result follow-up workflow, the ward board and a verified audit chain.

Transcript

The full narration, as spoken in the voice-over. Headings are the captions shown on screen.

Hospital data lives in feeds and systems that don't agree on who the patient is. Every new dashboard or assistant rebuilds the same plumbing, and none of it can say who saw what, or why. This is Thalix: a hospital AI platform in one on-site node.

It is built like a computer. Drivers turn every message, from HL7 and FHIR feeds to bedside devices, into one ordered event log that can be replayed. A kernel serves that log to people, apps and agents under one policy. Agents, workflows and live boards sit on top.

Every feed in one ordered, replayable log

This is a fictional demo hospital, fed by a simulator. In the console, admissions and lab results arrive over HL7, monitors over MQTT, and bed changes by webhook. A message is acknowledged only once it is safely logged, and every identifier links to one patient, so the systems agree on who this is.

Same patients, different views, one policy

In the workspace, a ward nurse sees Ward 4: beds, alerts, tasks, and each patient's results and timeline. A porter sees the same beds, but not names or results. Ask why, and the kernel names the policy that decided each part of the record.

The assistant proposes. A person confirms.

The assistant runs on a local model, on this machine. Asked what is waiting, it looks things up through the nurse's own access, and shows each tool it used. When it wants to change something, it proposes a card. Nothing happens until the nurse confirms it, and the action runs as her.

Workflows hand the work to the right people

Work between people and teams runs as workflows. An abnormal result starts a follow-up: a local model drafts two lines for the doctor, and the doctor records the outcome. A repeat test is a clinical order, so a consultant must approve it before it goes to the lab. Bed turnaround and discharge work the same way.

Live boards, resolved as the viewer

Boards on the canvas show live views, like this bed map, built by people or by agents, and resolved as whoever is looking.

Every read and action, hash-chained

Every read, search, action and denial is written to a hash-chained audit log, including what each agent and workflow did. The privacy officer can see who looked at what, and why, and check that the chain is intact.

It runs as one Rust binary inside the hospital, with the models beside it, on anything from a single PC to GPU servers. Sites without internet get an offline bundle.

Every demo runs on synthetic data. Thalix is not a medical device, and its agents never make clinical decisions: they prepare, check and propose, and people decide. Find out more at thalixhealth.com.

Read next

Thalix FAQ · HL7 v2 vs FHIR · AI agents in hospital operations · All guides

Thalix is at pilot stage. Everything in this video runs on synthetic data, and a hospital deployment needs its own clinical safety case, information governance approval and security review.

Thalix Health

See Thalix on a demo hospital.

A conversation with us is 45 minutes and free: the work you want to hand to agents, the systems involved, and where Thalix would and wouldn't fit. Demos run on synthetic patients.