Health digital twins, built with world models.
ari.ai replicates a person's body and health into a digital twin you can simulate. Our own foundation models go beyond diagnosis — they forecast what comes next.
Beyond diagnosis, into simulation.
Medical AI has spent a decade getting better at looking at what already happened. ari.ai is betting on the next step — AI that simulates what is about to happen. Replicate a body digitally, and the impact of a drug, a habit, or a treatment can be explored before it is tried.
Replicate
We fuse clinical, biosignal and lifelog data to digitally replicate a person's health.
Simulate
World models project the body forward in time, opening up the space of possible futures.
Decide
Those simulations help clinicians and individuals choose the right next move.
World Models — AI that learns reality.
A world model does not just classify patterns — it learns how the world unfolds over time. ari.ai has already shipped this paradigm in industry, and is now scaling it to the most complex world of all: the human body.
Foundation models that generate the next moment
We train our own large-scale generative foundation models from the ground up. Instead of one-shot predictions, they generate plausible futures across time, and let decisions be simulated on top of them.
That is fundamentally different from static classifiers or detectors. Being able to "pre-run" the consequences of an action is the core of what ari.ai builds.
We have done this in the wild before.
Before healthcare, ari.ai proved world models in esports. We built the world's first world-model-based generative foundation model for win prediction and strategy, and shipped it commercially to pro teams.
A world model for win prediction & strategy
A foundation model that simulates how a match will unfold moment by moment and recommends strategy on top of it — not a statistical layer, but a model of the game world itself. The first of its kind.
Sold to a professional team
We do not stop at research — we ship. The model was used in real coaching and operating decisions on the ground, and that experience is what gives us confidence to take this approach into healthcare.
For now, we are heads-down on the model.
A health digital twin is not a one-shot goal. ari.ai is concentrating resources on building the underlying world model itself — not chasing short-term revenue. The research depth of UNIST's Graduate School of AI and ari.ai's track record in industrial deployment sit on the same team.
Biosignal sequence learning
Foundation-model architectures that learn temporal dynamics from high-dimensional, heterogeneous data.
World-model-based simulation
Modeling that goes beyond observation — generating and evaluating plausible futures.
Clinical & real-world interface
Safely connecting model output to real medical and healthcare decision-making.
A team out of UNIST.
ari.ai is a healthcare AI startup led by a professor at the UNIST Graduate School of AI. Our research base sits at UNIST in Ulsan, with a Seoul office for industry and partnership.
Ulsan HQ
50 UNIST-gil, Eonyang-eup, Ulju-gun, Ulsan, Republic of Korea
Inside UNIST campus
Seoul office
TIPS Town, 165 Yeoksam-ro, Seoul, Republic of Korea
We are looking for people to build this with.
Research collaboration, clinical partnerships, careers, investment — reach out about any of it.
hello@ariai.io