Futuristic AI

Judge clinical AI with a spec, not a prompt.

Compile a Judge Spec, score a dataset, and check the judge against clinician labels before it ships.

Specs

—

—

—

—

Kappa per stored calibration report and mean score per run, oldest first.

The five newest eval runs for this tenant.

Three steps from a description to a calibrated judge. The seeded NHG demo already stores the note style guide under Corpora.

1Create a judge
Answer twelve questions or describe the task. Save the result as a draft.
New judge
2Add evidence
Upload a guideline corpus and import the rows you want scored.
Open corpora
3Run and calibrate
Score the dataset, then compare the judge with clinician labels.
Launch a run