Worked on the longhorn-manager repository to deliver support for fractional CPU values in the guaranteed instance-manager-cpu setting, enabling floating-point resource allocation for Instance Manager Pods. This feature improved resource granularity and scheduling accuracy in multi-tenant Kubernetes clusters, allowing for more efficient and predictable utilization. The technical approach involved updating the Types API to handle floating-point values, with careful attention to backend development and system configuration. Go was used throughout the implementation, leveraging strong type system skills to ensure correctness. The work focused on a clean, targeted feature delivery that addressed resource efficiency without introducing unnecessary complexity or changes elsewhere.