In this project, I developed AI agents capable of learning to play the game of Checkers using two distinct reinforcement learning algorithms: Deep Q-Networks (DQN) and Proximal Policy Optimization (PPO). The goal was to compare value-based and policy-based approaches in a moderately complex, strategic environment. Environment Setup: The game logic and board environment were implemented using custom Python code and integrated into a Gym-compatible format for training compatibility. The state representation captured piece positions, king status, and turn information, encoded into a structured input for the agents. DQN Agent: The DQN agent used a convolutional or MLP-based Q-network to estimate action values from board states. Experience replay and target networks were used to stabilize learning. PPO Agent: The PPO agent followed an actor-critic framework, learning a stochastic policy and a value function. It was optimized using clipped surrogate objectives to ensure stable policy updates. Training & Evaluation: Agents were trained in self-play mode, allowing them to iteratively improve by facing increasingly stronger versions of themselves. Performance was evaluated using win rates, average reward, and strategic depth of gameplay over time.