CMU PokerAI Competition

For Carnegie Mellon's competitive PokerAI tournament, I developed a sophisticated poker-playing artificial intelligence that combined machine learning techniques with classical game theory principles. The system employed a hybrid approach, integrating reinforcement learning algorithms with traditional decision-making frameworks to create a robust player capable of adapting to diverse opponents and game scenarios.

At the core of the AI were neural network models trained to evaluate hand strength and opponent behavior patterns. The architecture leveraged Monte Carlo simulations to explore potential game states and outcomes, enabling the AI to make probabilistically sound decisions under uncertainty. Advanced game theory optimization techniques, including Nash equilibrium concepts and exploitative play strategies, as well as random chance, guided the AI's betting patterns and bluffing mechanisms.

This comprehensive approach proved highly effective, with the AI achieving an impressive 73% tournament win rate and peaking on the leaderboards at 3rd place out of 70 competing teams from across the university.

Back to Projects