Private clinical documentation built around behavioral-health nurses.

SecureScribe is a clinical documentation AI project founded and created by Anson Washeck in Sacramento, California. It turns narrative nursing observations into structured, review-ready documentation using a compact model designed to run on facility hardware. The technology originated as Veridian, Anson’s earlier clinical AI project, which received recognition at Sacramento State University’s Hornet Hacks competition before evolving into SecureScribe.

• Workflow: document naturally, validate required elements, assemble the draft locally, and require nurse review before it reaches the existing EHR.
• Initial behavioral-health workflows include PRNs, refusals, behavioral observations, milieu events, safety events, and family contacts—the documentation that becomes difficult at the end of a long clinical shift.
• Safety is enforced through source grounding, missing-field flags instead of guesses, mandatory nurse approval, and no cloud AI inference.
• The project connects model evaluation with an operational pilot: nurse-adjudicated ground truth, held-out records, deliberate omissions, missing-field recall, matched prompting, and comparison against larger frontier models.
• SecureScribe is one of Anson’s primary projects and reflects his Edge AI focus: a small model, a constrained device, and a workflow designed around a real healthcare user.