AgentCon 2025 SUPERWISE Live Demo

Guardrails block or strip PII and PHI before they reach the LLM. In the demo, a medical guardrail agent sits between the user and the model so patient-identifiable data never touches the model. SUPERWISE enforces that at the platform layer.

Separating general medical knowledge from live patient records lets you answer clinical questions without exposing PHI. SUPERWISE connects both to the agent while guardrails ensure patient data never leaks into answers or training.

A simple assistant connects to medical knowledge and synthetic patient records. Queries like “common symptoms of asthma” or “how many patients have hypertension” are answered safely. With real data you must use a privately hosted model and HIPAA-aligned setup.

Medical data leakage breaks HIPAA and trust. SUPERWISE guardrails and auditability ensure that only safe, compliant outputs leave the system and that every decision is traceable for compliance.