An adaptive fuzzy control model is developed as an optimization-based real-time insulin delivery system for managing diabetes. The fuzzy controller uses 3 input values (blood glucose levels, rate of change of glucose, and amount of food consumed) to adjust the rate at which insulin is infused based on how close the glucose level is to the desired range. In comparison to traditional Proportional Integral Derivative (PID) type controllers, that have difficulty dealing with both time delay and the non-linear relationship of glucose and insulin, the adaptive fuzzy controller performs optimization by adjusting linguistic rules and tuning membership functions in real time to produce less erratic and more stable responses. The simulation demonstrates that the fuzzy controller is capable of reducing glucose excursions, preventing hypoglycemia, and returning to target more quickly than the PID controller after a meal. Comparisons of the fuzzy controller to the PID controller show the potential of this optimization-based technology as part of a closed loop artificial pancreas.