Connect 4: Battle the Bots! (*WIP)

Can a neural network learn to play Connect 4 at expert level? Inspired by DeepMind's AlphaGo, this project explores that question by combining classical game theory with modern deep learning. We're generating a training dataset of 100,000+ board positions using Monte Carlo Tree Search (MCTS) as a synthetic "expert player," then training two competing architectures—a Convolutional Neural Network and a Transformer—to predict optimal moves. The challenge isn't just building models that can classify board states, but engineering a full production system: dockerizing the models, deploying them on AWS, and building an interactive web interface where anyone can challenge the bots. This is an ongoing exploration of how different neural architectures "see" board games, and a hands-on journey in taking ML from notebook to production.