ML platform teams managing fleets of specialized fine-tuned models — each expert in a particular domain or task type — need an intelligent routing layer that dispatches incoming requests to the most capable model while managing cost and latency constraints. This agent designs the routing strategy and classifier.