This research provides a new angle to solving the problem of balancing and sequencing hybrid assembly lines by leveraging the power of artificial intelligence. Our mixed-model assembly line environment involves using Deep Neural Network (DNN) and Reinforcement Learning (RL)in order to consider both balancing and task sequencing properly. Production history data such as task times, total number of stations and relations between tasks were obtained and modeled. A Deep Q-Network based reinforcement learning model was used to learn the optimal task sequence and assign tasks to workstations in real-time and subject to the least makespan and in line with maximum line efficiency. Similarly, DNNs were used for predicting the time it will take to process tasks and the shifting of tasks among stations. Numerical evaluation on real world data shows that the new approach effective reduces idling time, minimizes waiting time of task and enhances the production flow. A comparison with more conventional optimization algorithms like Geometric Algorithms and Simulated Annealing also emphasizes the compound gain in terms of optimization time and near optimal solution that is possible with the machine learning based approach.