BanditGames (AI Board Game Platform)
A full-scale, AI-powered digital board game platform built on hexagonal microservices, event-driven RabbitMQ messaging, and custom MCTS AI opponents.
Co-architected an AI-differentiated board game platform built with Hexagonal Architecture and event-driven communication via RabbitMQ. An Anti-Corruption Layer enables external games to register dynamically and emit asynchronous state events without coupling to core platform domain logic.
Engineered intelligent game AI powered by Monte Carlo Tree Search (MCTS) featuring dynamic difficulty adjustment based on player skill. Built the complete ML lifecycle pipeline—collecting self-play data, versioning datasets with DVC, tracking models in MLflow, and deploying move prediction and win-probability estimation models via FastAPI microservices.
Integrated a vector-backed RAG chatbot with cached responses for low-latency rule queries, secured the platform using Keycloak SSO, and implemented automated GitLab CI/CD pipelines with strict quality gates for automated deployments.