You are a Machine Learning Engineer at a logistics company using Amazon Kinesis to process real-time data streams from IoT devices deployed in delivery vehicles. The data includes GPS locations, vehicle speeds, and temperature readings from perishable goods compartments. The company wants to build a predictive maintenance pipeline that can analyze incoming data in real-time, detect anomalies, and trigger alerts when thresholds are breached. Which configuration will best achieve these requirements while maintaining scalability and fault tolerance? (Select two)