Machine Learning, Data Science and AI

Our research area unites machine learning, artificial intelligence, and data science to build systems that learn from data, handle uncertainty, and act responsibly in the real world. Probabilistic and Bayesian methods provide the foundation for robust inference and decision-making, supporting reliable large models and reinforcement learning. We advance speech and language technologies—from modeling speech production and recognition to synthesis and watermarking—while safeguarding privacy and ethical use. We also develop deep and generative models, including transfer, meta-, and continual learning, scalable and federated training and post training, sensor fusion, and trustworthy generation for applications from scene understanding and robotics to AI enabled scientific discovery.

A central focus is AI for health and science: methods that aid clinical decisions, reveal molecular mechanisms, and model complex dynamical systems in physics and biology. Examples include computational biomedicine, digital phenotyping for mental health, and data driven modeling to accelerate experiments. In parallel, our work in computer vision and multimodal learning teaches machines to interpret visual data and combine it with text, audio, and sensors, powering autonomous perception, vision–language understanding, and scientific imaging. Across all areas we embed fairness, transparency, robustness, safety, and explainability, and contribute to standards and evaluation—combining theory with practice to make AI more capable, trustworthy, and beneficial.