Whenever you're ready, walk me through how you'd approach this.
Describe a time you used an orchestration framework like LangGraph to solve a multi-step data processing problem for an Autodesk-style AI workflow.
Explain how you modeled task dependencies and shared state, handled validation, retries, idempotency, partial failures, and branching, and monitored the workflow from ingestion through output delivery. Include the technology choices, representative implementation details, and how you measured pipeline reliability and data quality.