Toward Trustworthy Clinical AI: Building Safe & Context-Aware Systems for Healthcare
Date:
📄 Slides
Invited talk on building trustworthy clinical AI, covering:
- Routing: getting the right question to the right specialist (eConsults → SAGE)
- Harm: evaluating clinical safety of AI recommendations (NOHARM)
- Messaging: detecting & preventing errors in AI-drafted patient-portal replies (retrieval-augmented guardrails)