Structured and unstructured records hold the answer to who qualifies, but they sit inside the site's walls — where the AI has to go, not the other way around.
Manual chart review cannot cover eligibility signals hidden in unstructured notes and imaging at scale, so screen-failure rates stay high and timelines slip.
Clinical development, trial operations, feasibility, and data science teams, with academic and hospital study sites.
Pre-screening runs against de-identified records at the site or inside a sovereign deployment you control — sovereign cloud, on-premise, or air-gapped. Source data is never required to leave the environment it already lives in for the model to read it.
Inclusion and exclusion criteria become machine-checkable rules with the ambiguities flagged, not guessed.
Generative models read de-identified notes, labs, and history to surface candidates with the evidence cited.
Segmentation refines the shortlist against imaging endpoints — OCT today, extensible by modality.
Cohort counts per site before commitment, so recruitment plans rest on data rather than optimism.
No. Pre-screening runs on de-identified records at the site, or in a sovereign/on-premise deployment you control. Only de-identified, permitted derivatives move, for multi-site feasibility rollups.
Protocol criteria are translated into machine-checkable rules with ambiguities flagged for a clinician to resolve, rather than silently guessed by a model.
No. The platform produces a clinician-reviewed shortlist with the evidence that triggered each match cited. A human always makes the final eligibility call.
Yes. Cohort counts per site are produced before commitment, so recruitment plans rest on data rather than optimism.
Our proprietary AI orchestration platform for healthcare: one engine, a registry of reusable task modules and domain agents, and a governed knowledge base — deployed inside your walls and run by your team.