You are a machine learning engineer at a fintech company with three models: fraud detection, credit scoring, and personalized marketing. Each model has distinct performance and deployment needs: the fraud-detection model requires real-time, low-latency predictions and rapid scaling with transaction volume; the credit-scoring model is computationally intensive but can run in batch with moderate latency; the personalized-marketing model is event-driven and does not require constant availability. Given these requirements, which deployment targets are MOST SUITABLE for each model?