Mahsa Bazzaz

I am a doctoral student at Northeastern University’s Khoury College of Computer Sciences, specializing in procedural content generation for discrete data types like game levels. My research combines traditional constraint-based models with modern machine learning techniques to generate diverse, playable, and computationally efficient game levels. My goal is to develop ensemble methods that capitalize on the reliability of constraint-based models alongside the speed of machine learning approaches. I am particularly interested in bringing together communities that rely on generative models to produce “correct” outputs like molecular design, arithmetic expression generation, and source code synthesis. These communities deal with underlying tasks that are highly similar to game level generation, therefore game level generation can offer an ideal test bed that avoids ethical concerns and excessive complexities for them.