Keeping blood glucose levels within the safe range is a persistent issue in Type-1 Diabetes (T1D) management that is especiallyprominent during periods of mealtime rapid glucose variability. Traditional insulin management approaches, including Proportional–Integral–Derivative (PID) and heuristic Bolus-Based (BB) algorithms, rely on rules that are more or less manually tunedand using fixed pre-meal dosing times. These rules often lead to suboptimal glucose control, resulting in either hyperglycemiadue to delay onset of action in injected insulin, or hypoglycemia from overdose. We present a meal-timed reinforcement learning(RL) algorithmic framework that jointly optimizes insulin bolus dose and timing for glucose control. Using a Soft Actor-Critic (SAC) agent, the control system learns adaptive dosing algorithms based on continuous glucose feedback, meal intake, andphysiologic variability. The role of RL was evaluated by training in three separate virtual patient groups—the adult, adolescent, and child patient groups—each of which represented 30 virtual patients per cohort for 10 days of simulation—totalingin 900 virtual patient days of evaluation. Results from the RL control system found enhanced glucose control compared toconventional PID and BB controls, maintaining average Time-in-Range (TIR) of approximately 80% while also reducing TimeBelow-Range (TBR), and reduction in excess insulin utilization. Overall, these results exemplify that when dose and timingare jointly optimized with adaptive RL-based approaches, glucose control is accomplished more safely and physiologically thanwith commonly accepted static rules.