Python expert for Reinforcement learning

You will implement the Q-learning algorithm, and evaluate its
results on a simple gridworld environment (see Fig. 4 with optimal values), which
is an episodic task with zero reward everywhere except for a reward of +1 when
reaching on the goal state/cell. Once the goal is reached, the episode is terminated.
Your implementation must be able to handle arbitrary numbers of (discrete) states
|S| > 0 and actions |A| > 0. The vectors and matrices are represented as NumPy
arrays. Your functions shouldn’t print additional information to the standard output.