My research focuses on understanding and designing representation learning frameworks that are generalizable, interpretable, and controllable. I aim to develop representations that generalize across diverse tasks and domains, are interpretable to support meaningful insights and decisions, and are controllable to enable direct manipulation for various applications.