Features
Daily signed governance PDF per role
Drift monitoring (PSI) with baseline comparison
Subgroup fairness disparity detection (HHS §1557)
Silent update detection for vendor AI models
Latency p95 vs. target per model
Radiologist thumbs-up/thumbs-down feedback aggregate
Alert management with acknowledgment tracking
Evidence packets with chain of custody
Offline report verification with public key
Continuous posture monitoring (ARCH-AI, Assess-AI, ONC HTI-1, FDA PCCP)
IHE-aligned (UPS-RS work-list status)
HIPAA-safe egress (no PHI on wire)
Multi-vendor AI ingestion via HL7 and DICOM
Role-based report (IT Director, CMIO/CIO, Compliance Officer)
SageMaker AI for full ML lifecycle (build, train, deploy)
SageMaker Unified Studio for analytics and AI development
Amazon SageMaker Catalog for data and AI governance
Lakehouse architecture unifying S3, Redshift, and federated sources
Zero-ETL integrations for near real-time data ingestion
HyperPod for distributed training of large models
JumpStart for pre-built models and solutions
MLOps tools for model management and versioning
Amazon Q Developer for natural language productivity
Federated query across third-party data sources
Fine-grained access controls and permissions
Built-in data quality monitoring and lineage
Apache Iceberg compatibility for open table formats
Bedrock integration for generative AI applications
Web Search on Bedrock AgentCore for grounded agent responses
