My broader research topic is the design and/or learning of learning priors for deep neural networks, such that systems that are robust to transfer across data domains, tasks and modalities is possible. To achieve this, I utilize meta-learning, differentiable and non-differentiable learning, adversarial learning and multi-modal learning, which is where the majourity of my current work revolves.