Home

Projects

In Progress · 2024 – Present

Neural Network Chess Engine

Competitive AI chess with a hybrid classical + neural architecture

A competitive chess engine achieving 2000+ rated play by integrating classical search algorithms with neural network-based evaluation. This hybrid system combines the strategic depth of traditional minimax algorithms with the pattern recognition capabilities of deep learning. Built a hybrid system interfacing optimized C++ move generation and alpha-beta pruning search with a PyTorch neural network for position evaluation, trained on thousands of expert and self-play games, with board representation optimized using bitboards and Zobrist hashing.

Features

  • Achieved 2000+ rating in competitive play
  • Efficient move generation using bitboards
  • Neural network trained on expert game data
  • Optimized search with alpha-beta pruning

Tech Stack

PythonC++PyTorchMinimax SearchAlpha-Beta PruningBitboardsZobrist Hashing