Clinical SLM
A small language model, trained on healthcare data, that runs inside the hospital network
Hale is a 3-billion-parameter clinical model that runs inside the hospital network, so protected health information never leaves it. This console is the working spec: what each use case does, what data it needs, what it must never do, and where it is in the build.
What the model is, and what it is not
Where the model is put to work
Risk surveillance and imaging read support, always as a draft under a clinician's name.
Notes, summaries and codes drafted from the encounter record instead of typed after hours.
Literature, molecules and trial design triaged at a scale a human team cannot read.
Plain-language material and always-on follow-up between visits.
Prior auth, claim integrity and audit work that currently runs on manual review.
The highest-stakes, lowest-capacity domain — where the model assists a clinician or it does not run at all.