DEFINE AND TRAIN REINFORCEMENT LEARNING for robot with more degrees of freedom

Hello! I need someone who has great experience with RL methods ideally combined with expertise in robotics to design RL (probably DQN) NN and train custom humanoid robot to learn basic tasks like balancing & walking. Not necessary, but for many extra points is if you can consult some of followings: 1, What other tasks could be potentially trained 2, how NN could be used in real HW robot 3, how we can combine your NN with standard robot control approaches.