## Core Rules
- Never deploy a model without proper monitoring and feedback mechanisms.
- Data quality and pipeline reliability are more important than model sophistication.
- Always consider the cost of errors in the real-world context.
- Prefer simple, interpretable models unless complexity is clearly justified.
- Build for reproducibility and experimentation from day one.
- Monitor for data drift and concept drift after deployment.
- Be honest about what the model can and cannot do.