True care always listens
AI-augmented doctors & practice. A copilot built for medicine, designed on the ground in a working primary-care practice in Luxembourg. Local, GDPR-by-construction AI that gives clinicians their time back, and never takes the decision from them.
Handles the data, the documentation and the noise at the moment it appears, so the clinician gets the time back to listen and to reason.
Scheduling, waiting room, incoming documents, follow-ups, billing. The layer a primary-care practice actually runs on, with the same engine underneath.
Pre-visit context and patient-reported data, under the patient's control, with explicit choice over what is shared. No diagnosis, ever.
Biology, lab reports, faxes, mail, any source. A local, purpose-built AI extracts what matters and files it in the right patient folder automatically. The doctor decides what happens next.
This is the piece already running daily in the practice, and it is our proof that the hard part is not the model, it is the last metre into the clinician's workflow.
Local AI means GDPR by construction, not by paperwork: clinical data is processed on site, not shipped to a third-party model.
Parsed, coded and filed under the right patient, with values presented in context.
Correspondence attached to the episode of care it belongs to.
Still the reality of primary care. Read, classified, routed.
Everything filed and searchable, so the clinician reviews a clean inbox instead of a backlog.
Not one monolithic model trying to know everything. Medicine reasons by specialty, and so do we: each organ system gets its own focused expertise, cross-checked against the others. That mirrors the very nature of polypharmacy, which is an inter-organ problem.
NeuroCardioHepaticRenal
Gastro
Immuno
BonePatient modelThe frontier moved from one fine-tuned model to systems that reason in parts and check each other. We built for that frontier, and architected for the one after it.
We know how unreliable free-text models are, so we don't ask for prose. Each step returns a bounded, typed answer with an explicit confidence level: below threshold, nothing is surfaced as a finding, it goes to the clinician. Reliability comes from the constraint, not from hoping the model behaves.
Symptems produces structured signal for the clinician to act on. The clinician remains the decision-maker and the liability holder, surfaced explicitly in the interface.
An iatrogenic cascade has a cause and it has dates. That ordering is the signal every current tool throws away, and it is where our architecture has an edge.
A first medication is prescribed and tolerated.
A second medication inhibits the enzyme of the first.
The phenotype shifts. Nothing visible yet.
The adverse effect appears, too late to be attributed to the right culprit.
A third medication is prescribed for that symptom.
The genotype is static; the phenotype is a trajectory, and therefore computable by an event-driven engine. Our exploratory work on masking mechanisms points this way. It is a lead, not a demonstration, and we are building the clinical partnerships to validate it properly.
We build inside an active general-practice clinic at the heart of the Greater Region in Luxembourg, with real patients and real constraints. That is the difference between designing decision support and discovering, late, that it does not fit the consultation.
Luxembourg is a multilingual country with a pilot-scale health system and direct access to GPs in both initial and continuing training: a short channel to recruit, and a short channel to reach the market.
It also carries a population where a large share lives on one side of the border and is treated on the other. Prescriptions accumulate across systems that do not yet talk to each other, which is exactly the gap European interoperability is meant to close.
Small enough to move, real enough to prove. Direct access to GPs in initial and continuing training.
Clinical software and patient interaction tested across languages from day one, not retrofitted for expansion.
Patients treated on both sides of the border, with medication lists no single system holds in full.
One clinical site, one channel to recruit, one channel to disseminate. Real-world constraints from the start.
Dr Eric Sassel is a practising primary-care physician and a member of the Luxembourg medical digitalisation committee, working on making the country's medicine coherently and interoperably digital.
That places the product where it needs to be: close enough to the workflow to design it right, and close enough to the regulation to build for where European healthcare is actually heading.
Symptems is built in his own practice, with real patients and real constraints, which is why the hard part we solve first is the last metre into the clinician's day rather than the model itself.
Primary-care physician and founder. Member of the Luxembourg medical digitalisation committee (EHDS). Brings the clinical reality and the national health-data view straight into the product.
Neuropathologist of international standing, co-author of the WHO classification of tumours of the central nervous system.
Engineer specialised in multi-agent systems and data security, after deliveries for Fortune 500 companies and large public organisations.
We are opening the data room to a small group of partners. If healthtech at the intersection of clinical reality, AI architecture and European regulation is your field, we would like to talk.
So the right person gets back to you, with the right thing.
We read every message ourselves and usually reply within a couple of days.