Kubernetes, a container orchestration platform available as open source, ensures high availability and scalability by employing various autoscaling mechanisms such as Horizontal Pod Autoscaler (HPA), Vertical Pod Autoscaler (VPA), and Cluster Autoscaler (CA). Playing a pivotal role in ensuring a smooth service experience, HPA dynamically adjusts the number of replicas (pods) by scaling up or down. Kubernetes actively monitors essential Resource metrics, such as CPU and memory usage for pods. This study explores HPA through a series of diverse experiments to offer valuable insights into its operational behaviour across various machine learning (ML) applications. We analyze and provide insights into the behavior of HPA across various machine learning model based applications. Lastly, we delve into the significance of custom metrics collection for specific ML applications, offering deeper insights and lessons learned.