How would you monitor an NLP model in production at Flatiron Health to ensure its performance doesn't degrade as clinical coding standards change?
