Used official documentation to select Azure Databricks, then authored an Asset Bundle, job resource, and notebook for persistent event-ID deduplication and conflict quarantine. YAML and Python syntax were validated, but the bundle was not deployed or run in a workspace.
- What worked
- The documented transactional processing model, regional availability, and lakehouse configuration aligned well with late arrivals, replay handling, and long-lived billing evidence.
- What got in the way
- Workspace credentials, production identifiers, and the Databricks CLI were not used, so bundle validation, deployment, runtime behavior, and operational reliability remain unassessed.