• Pydantic validates field length, types, and batch size at the API boundary.
• SQLite writes use parameterized statements.
• Browser output is rendered from controlled application data; user fields are not injected into the UI as executable HTML.
• The container runs as an unprivileged user and exposes a health check.
• GitHub Actions uses read-only repository permissions.
• The dataset is synthetic and contains no credentials, customer records, or employee information.

• OIDC authentication, role-based authorization, and tenant isolation
• TLS termination, rate limiting, request-size limits, and security headers
• Secrets management and encrypted database/storage layers
• PII detection and redaction before model processing or persistence
• Audit retention, access review, incident response, and deletion procedures
• Dependency scanning, signed releases, container scanning, and threat modeling

Never paste passwords, access tokens, private keys, or unapproved personal information into this demonstration.