SURF Projects

Undergraduate research exploring procedural content generation using diffusion models and large language models (LLMs).

Mario Diffusion

Trained diffusion models with UNet architecture for procedural content generation (PCG) in game level design, with optional text-based guidance. Contributed to model infrastructure (early stopping, checkpoint resumption), improved mixed-initiative tools for level building, and developed scripts to evaluate and visualize runtime and A* solvability metrics.

LLM and Stable Diffusion Story Game

Built a real-time multiplayer storytelling game using Flask, JavaScript, and Socket.IO. Implemented UI logic, LLM output handling, and a goal-based win condition where players steer the narrative toward secret, personalized endings evaluated by the LLM.

SURF Reflection Essay

Final reflection on my time during SURF, including project descriptions, acquired skills, and personal takeaways.