Independent evidence and testing lab for AI in healthcare
trust.
creiamo.ai tests whether the AI inside a healthcare product does what its maker claims, and meets what its regulation requires: UK MHRA, EU Medical Device Regulation (MDR) and AI Act, FDA, Health Canada.
AI makes healthcare better only if it is trusted.
AI can make healthcare better for all. For that to last, it has to be precise, trusted, and within the mandate it was given. Probabilistic systems make that mandate hard to prove. We build the evidence that proves it.
Deterministic
Same input, same output. Test once, sign off, ship. The assurance holds because the system cannot surprise you.
Probabilistic
Novel outputs from identical inputs, and behaviour that evolves after deployment. The AI-specific evidence, robustness, drift and evaluation, is the newest part of the file, and the part reviewers question most.
* Team-NB, Medical Device Survey 2025 · team-nb.org
The AI Act (Digital Omnibus) high-risk obligations have imminent deadlines. A review may take 13 to 18 months.
Five products. One published method.
Every report follows the Evidence Profile: a public, citable method for testing AI against its declared behaviour and the regulation that applies to it. You decide whether a report becomes public, and when. Each product returns evidence you own, can share with a reviewer, and use to improve the AI we looked at.
Evidence for certification.
Your AI documentation and supporting artifacts tested against the Evidence Profile for your route. Robustness, drift characterisation, evaluation under stress and edge-case conditions. A report a notified body or reviewer can assess, carrying our name, the method version and the date.
Test before you submit.
The same tests, on the same method, run on a candidate release before you commit to a submission. How the model holds up on noisy or degraded inputs. Results broken down by subgroup. Drift between the training data and the population you serve. Behaviour on inputs outside the intended use. What changed between one model version and the next.
Monitoring after the certificate.
The product instrumented, so post-market surveillance (PMS), post-market clinical follow-up (PMCF) and periodic safety update report (PSUR) inputs flow from a live evidence stream and certified behaviour is checked continuously, not retrofitted under change control. As the product is updated or taken to new markets, the evidence stays up to date.
Evidence for the deployer.
For the hospital or health system buying or running AI. Does it still do what its file says on your data, in your population, in your workflow. And does it stay that way.
Enablement, not consulting.
Enablement helps you get the most from the products. It is an add-on, priced separately.
Onboarding and integration. Training on the Evidence Profile. Interpretation of results — what the evidence says, how a reviewer will read it. Set-up of ongoing monitoring.
Who it is for — and who pays.
Those who create or deploy AI in healthcare pay for the evidence. Those who assess it never do.
Developers of AI in healthcare
Software as a medical device (SaMD), AI inside a medical device, AI-as-a-medical-device. Evidence for submission, tests you run yourself, and monitoring once you are on the market.
Request a scoping callHospitals and health systems
You buy and run AI. Evidence that it does what its file says on your data, in your population, in your workflow, and that it stays that way.
Request a scoping callRegulatory consultants
You assemble the technical file. We are the lab behind your practice: evidence to your scope, under your review. You keep the client, the file and the relationship.
Work with usFour rules we will not trade for revenue.
An evidence lab is worth exactly as much as its independence. These four rules protect it, and they are not negotiable in any engagement, at any price.
The method is public.
The Evidence Profile our reports follow is published and citable by anyone: a client, a reviewer, a competitor. Part 1 opens for consultation as a dated draft. Coming soon.
Every report is signed and dated.
Each one carries creiamo.ai's name, the method version and the date. A client can decline to use a report. It cannot alter or suppress it.
No money crosses the review.
Reviewers and notified bodies never pay creiamo.ai, and creiamo.ai never pays them. Whatever we give them, including enablement, is free.
The price is fixed and public.
It is stated before we start, it is the same for everyone, and it never depends on the result. The public price list is released when the Founding Partners Program ends.
Frequently asked questions.
Is creiamo.ai a notified body?
What does it cost?
Which markets does creiamo.ai work in?
Hold the evidence before the review begins.
For developers of AI in healthcare (SaMD or AI inside a device) heading toward, or already holding clearance, and for the hospitals and health systems that deploy them. Paid at a discounted founding fee: fixed, stated before we start, the same whatever the result, and fully credited against the first-year service if you continue.