Design automated anomaly detection and remediation for cloud services

Design a platform that watches metrics and logs across many cloud services, detects anomalies, and responds automatically where it is safe (restart, roll back, scale) while paging people otherwise. Cover the detection models and their thresholds, avoiding false alarms, the workflow that runs remediations, guardrails, and learning from past incidents.