This project presents the training of a Doom game agent using reinforcement learning. Reinforcement learning is used to train the agent over no. of episodes using vizdoom which is a visual environment that allows us to play doom using python and other programming languages. Vizdoom then is used in open Ai gym environment to build machine learning model. Our model is trained using PPO (proximal policy optimization) algorithm in various scenarios over no. of episodes. As no. of episodes are increased and the agent is trained it can be seen that mean reward increases and our agent performs better actions each time.