Project
AI patient-intake platform
LLM features engineered for HIPAA-regulated data
An AI patient-intake platform for U.S. healthcare. The hard problems here are trust problems: letting a large language model help with sensitive clinical workflows without ever putting protected health information at risk.
I architected the platform and led its build. Its LLM features run through a resilient gateway to Anthropic Claude on AWS Bedrock, with prompt-level PII stripping and circuit breakers. Protected health data is encrypted at the field level with managed key rotation, behind clean, auditable boundaries. HIPAA was the floor, not the ceiling.
