Inquiry Response: Safeguards For Frequently Retrained Models In Production
We deploy an automated, predictive service prioritization model every week, training on the past four months data and refreshing the previous week’s data (subtracting the oldest week). Is this frequency justified? Also, what safeguards should we have in place?
THE BIG IDEAS:
- Safety net
- Manual tasks
You have an impressive MLOps capability in terms of model deployment, but there’s a statistical component to MLops as well.
- As a best practice, the person