The two sides of clinical AI, the research that produces it and the practice that lives with it, are built in different geometries. Pendentiv exists because we've spent our careers on both and watched the same failure repeat from each side of the wall.
Most clinical AI efforts are anchored at one end: brilliant research that never survives contact with a workflow, or slick deployment of models nobody should trust. The partnership is the bridge, and each discipline keeps the other honest.
Close to a decade building, deploying, and maintaining machine-learning models inside healthcare: production systems, health-system data, integration constraints, and the operational reality of keeping models alive and trustworthy after go-live.
Academic rigor rooted in decision-making analysis: judging clinical AI by the decisions it changes rather than the benchmarks it clears, grounded in years of research on how clinicians weigh evidence at the point of care.
A pendentive is the curved triangular structure at the corner of a square room that carries the geometry upward, from the flat walls below to the round base of a dome above. It doesn't hold the dome. It makes the dome possible on a foundation that otherwise couldn't bear it.
Machine learning is built in one geometry: research, curated data, controlled conditions. Clinical care lives in another: variable populations, embedded workflows, provider judgment, institutional culture. Pendentiv is the structural element between them.
The commitments underneath every engagement — the ones we'd rather lose work over than bend.
Nothing reaches a clinician until the evidence holds for their patients, their setting, and their standard of care. "It worked elsewhere" is a hypothesis, not a credential.
The work happens where care happens. Workflows, decisions, and trust can't be mapped from a conference room in another time zone.
Clinicians shouldn't reorganize care around software. If adoption requires willpower, the design has already failed.
Clinical judgment is the point, not the obstacle. Where stakes are high, review is built in — and the clinician always has the last word.
Sustained use, honest override counts, real clinical impact, reported the way we'd report a study, including when the answer is "stop."
Every engagement is documented to publication grade. Our partners' papers, grants, and grand rounds matter as much as the go-live. The work has to stand up in both worlds.
Not a strategy consultancy. Not a research lab. A build shop with a research thesis, embedded where care happens: the one vantage point where the thesis can be tested.
Tell us about your models, your setting, or the gap you keep hitting. We'll bring both sides of the bridge.
Work with us hello@pendentiv.ai