This paper addresses the challenge of robust and adaptive control for lower-limb exoskeletons operating under nonlinear dynamics and user-dependent parametric uncertainties. Human variability in parameters such as body mass, joint stiffness, and gait characteristics significantly degrades the performance of fixed-parameter controllers, often requiring costly re-tuning for each user. To address these limitations, a hybrid AI-augmented adaptive control framework integrating Model Reference Adaptive Control (MRAC) with Deep Deterministic Policy Gradient (DDPG) is proposed. The resulting DDPG-Augmented MRAC (DAM) controller combines the stability guarantees of MRAC with the learning and generalization capabilities of reinforcement learning, enabling improved robustness and inter-subject adaptability. The proposed controller compensates for parametric variations across different users without requiring system redesign or full retraining. Simulation results demonstrate enhanced trajectory tracking accuracy and robustness under varying biomechanical conditions, indicating the potential of the proposed approach for intelligent and personalized control of wearable robotic systems.