Developed and delivered a memory-based autoscaling feature for the admission controller in the kyverno/kyverno repository, expanding the existing CPU-based scaling mechanism to also respond to memory pressure. This enhancement involved integrating new metrics hooks and scaling logic using Kubernetes and Helm, enabling dynamic adjustment of pod replicas during memory spikes. The implementation, written in YAML, addressed the need for reduced admission latency and improved cluster stability under fluctuating workloads. By providing clear ownership and targeted commits, the work improved the reliability and responsiveness of policy admission workflows, contributing to more efficient resource utilization and overall cluster throughput in cloud environments.