// PROJECTS

Cool stuff.

checkers mcts

[python][ai][game theory]

I wrote checkers from scratch in Python and built two AI agents to play it: a minimax agent that looks a few moves ahead, and a Monte Carlo tree search agent that plays out hundreds of random games to see which move wins most often. With 1,000 simulations per move, MCTS beat a random player in 100 out of 100 games. Against depth-3 minimax it won only 5%, drew 40% and lost 55%. The board here is a smaller MCTS playing live in your browser.

>> code on github >> read the pdf

pac-man search

[python][ai][search]

Teaching Pac-Man to find his way around a maze. I implemented depth-first, breadth-first and A* search, plus heuristics for harder problems like touring all four corners or eating every dot on the board, and proved my food heuristic is admissible and consistent. The maze here cycles through the searches: shaded squares are everywhere each one looked before finding the dot. (Scored 26/25 for my intro to AI class.)

>> code on github >> read the pdf

>> contact me