Application created using Unity and ML-Agents to explore the generation of dungeon-style levels using reinforcement learning. The aim of this university project was for the AI to learn to place a number of enemy and health pick-up tiles within set ranges that specify the 'optimal' level state. The report linked below conducts an investigation into the effect reward shaping has on the AI, as well as the effectiveness of both Proximal Policy Optimisation (PPO) and Soft-Actor Critic (SAC) trainers.