Any Equipment, Always Improving — How EMBRY and AT Lab Free Your Clinic From Vendor Lock-In

IVF Lab

Walk into ten different IVF clinics and you will find ten different setups. Different incubators. Different microscopes. Different time-lapse systems. Different software, different cameras, different file formats.

And yet, almost every AI tool on the market expects you to use exactly their hardware to get any value out of their software. Worse, the model that comes bundled with that hardware is usually frozen the day you buy it.

EMBRY was built on the opposite belief — that your hardware should not decide whether you get to benefit from modern AI, and that the AI you do use should keep getting better as long as your clinic does.

The Problem With Bundled AI

Many time-lapse and imaging systems ship with their own AI models pre-installed. On paper, that sounds convenient. In practice, it creates two big problems.

The first problem is cost. Premium hardware with bundled AI is expensive. Smaller clinics, partner sites, and clinics in emerging markets often cannot justify the price tag — so they go without AI altogether. The technology that could most help them is the technology they cannot afford.

The second problem is staleness. Most clinical AI on the market today is what regulators call a locked algorithm: it ships trained on the vendor's dataset and stays exactly that way until the vendor decides to release a new version, often years later. It does not learn from your cases. It does not adapt when your protocols change. Your clinicians can correct its predictions a thousand times and the next patient still sees the same answer. You are tied to the vendor's roadmap, not your own.

That is not an accident. Updating a regulated AI model traditionally meant a full new submission to the regulator — slow, expensive, and worth doing only every few years. So vendors took the easy path: lock the model and ship it. The clinic absorbs the cost of that decision in the form of an AI that quietly drifts out of sync with how the lab actually works.

EMBRY Works With What You Already Have

EMBRY is designed to be hardware-agnostic from day one. It connects to your existing equipment instead of asking you to replace it.

Whether your clinic uses:

• A premium time-lapse incubator from one of the major vendors

• A standard incubator paired with a separate microscope and camera

• An older imaging setup that simply produces image files or video

• A mixed environment where different rooms use different equipment

EMBRY pulls the data in, normalizes it, and runs the same modern AI pipeline on top of it. The clinic decides what hardware fits its budget and workflow. EMBRY adapts to that choice — not the other way around.

That single design decision has a real consequence. A small clinic with a modest setup gets access to the same AI capabilities as a flagship lab running top-of-the-line equipment. AI stops being a privilege of the well-funded.

Bringing Bundled Models Forward

Many clinics already own equipment that came with its own AI model — and they would understandably like to keep using what they paid for. EMBRY does not force you to throw any of that away.

EMBRY can run alongside vendor-bundled models, compare their predictions to AT Lab models trained on your own data, and let your team decide which to trust on a case-by-case basis.

Over time, most clinics see the same pattern. The bundled model holds its own on the cases it was originally trained for, but the AT Lab model — the one that learns continuously from your clinicians — pulls ahead on everything else. Edge cases. New protocols. Patient populations the vendor never saw.

You do not have to make a single hard switch. You can run both, measure both, and migrate at your own pace.

Continuous Improvement, Done the Right Way

Letting clinics retrain their own models sounds like a regulatory minefield — and it would be, if it were done casually. The reason vendor models stay locked is not laziness. It is that every uncontrolled change to a clinical model is a real safety risk and a real compliance problem.

So AT Lab is not just a "retrain whenever you feel like it" tool. It is the framework that makes continuous improvement safe and defensible:

• Controlled training protocols — every retrain follows a predefined process. What data is eligible, how it is reviewed, what evaluation gates a new model must pass before it can be deployed.

• Validation before promotion — no new model goes live without measurable proof that it performs at least as well as the current one, broken down by category and patient profile.

• Full traceability — every model version is linked to the exact dataset, annotators, and training run that produced it. If anyone asks how a prediction was made, you can answer.

• Reversible deployments — a model that underperforms in production can be rolled back instantly. The previous validated version is always one click away.

• Audit-ready logs — annotation, training, evaluation and deployment events are all recorded, signed and retained, so your team always has a clear record of how each model came to exist.

• 
  

This matters because clinical AI cannot be treated like consumer software. Every change has to be defensible. AT Lab is designed around that reality — so a clinic that retrains regularly is not a clinic taking shortcuts, it is a clinic operating inside a controlled, transparent process that mirrors the way good clinical work is done in any other part of the lab.

The result is the best of both worlds: your clinic gets a model that keeps learning, and you get the validation, traceability and control that a clinical environment requires.

Always Up to Date — Regardless of Hardware

This is where the architecture really pays off. With a vendor-bundled model, every new capability is gated behind a hardware upgrade or a paid software release. With EMBRY and AT Lab, new features arrive as platform updates — and they reach every clinic, on every supported hardware setup, at the same time.

When AT Lab adds a new training capability, every clinic gets it.

When a new evaluation metric ships, every clinic gets it.

When the model architecture improves, every clinic gets to retrain on the better foundation — inside the same controlled process they already use.

The hardware your clinic happens to own does not gate the software it can run. A ten-year-old microscope and a brand-new time-lapse incubator can both feed into the latest AT Lab pipeline and benefit from the latest models.

One Standard Across a Mixed Fleet

This becomes especially powerful for organizations running multiple clinics — exactly the case we covered in our previous article on cross-clinic model sharing.

Most multi-site networks did not buy all their equipment at once. The flagship site might run premium hardware. The newer satellite clinics might run something more modest. A recently acquired clinic might be on a completely different vendor.

Without EMBRY, that fleet is fragmented. Each site is locked into its own vendor's tools, its own data formats, its own AI capabilities. Reporting is inconsistent. Quality control is inconsistent. Patient experience is inconsistent.

With EMBRY in place, the entire network runs on a single software layer regardless of what is sitting on the bench. The same dashboards. The same models. The same metrics. The same continuous improvement loop — operating under the same controlled, audit-ready process at every site.

Hardware becomes a budget decision, not a strategic one.

Why This Matters

The history of healthcare technology is full of expensive systems that locked clinics into a single vendor's roadmap. Buy the hardware, accept the model, hope for an upgrade in three years.

That model does not work for AI. AI improves on a timescale of weeks, not years. A platform that can only move at the speed of your hardware vendor's product cycle is a platform that is always behind. And a model that cannot be safely retrained is a model that gradually stops reflecting how your clinic actually practises.

EMBRY decouples all of it. Hardware is hardware. Models are models. Hardware is hardware. Models are models. Rigor is built into the process, not bolted onto a frozen artefact. Each layer can evolve independently, and your clinic gets to benefit from progress on any of them without waiting for the others.

That is what makes modern AI accessible — not just to flagship labs, but to every clinic that wants to give its patients the best possible care with the equipment it already has, and to keep getting better at it every month.

Get Started

If you are not sure whether EMBRY will work with your current setup, the answer is almost certainly yes. We have integrated with a wide range of incubators, microscopes, and imaging systems — and we are continuously adding more.

Tell us what you have, and how often you wish your AI could learn from your team. We will show you what is possible.

Contact us to schedule a demo.