We are a group of data-driven machine learning enthusiasts who are primarily interested in building unified approaches to integrate and learn from complex real-world data. Beyond just text and images, we also build novel deep learning models that consider graphs, time series, geometry and tabular data, and use them to solve a wide array of applications in domains such as biology, medicine, chemistry, physics, neuroscience, social networks, science of science and supply chain.

Motivated by real-world use cases, we focus on efficient and scalable techniques that combine relational reasoning, multimodal learning, geometric deep learning and foundation models. Furthermore, we are actively doing research in trustworthy deep learning to allow safe, transparent and reliable deployment of such models.