It forecasts future carbon intensity values for the grid

It classifies incoming workloads as flexible or inflexible based on their resource requirements

It measures the actual energy consumed by each job after execution

It holds deferred flexible jobs until a lower-intensity execution window becomes available, then releases them