Students will participate in projects in one or more of the following research areas:

• Fundamental AI/ML: The team for fundamental AI/ML research focuses on developing novel AI/ML methodologies in order to address grand mathematical challenges, e.g., high dimensionality, lack of robustness, uncertainty quantification, etc., arising from scientific applications at DOE experimental and high-performance computing facilities.
• Scientific Imagery/Image Analytics: At ORNL we are tackling scientific deep learning applications in medical imaging, manufacturing, and geospatial AI. Our projects are pushing the state of the art in 2D and 3D image segmentation, classification, registration, and tomographic reconstruction leveraging ORNL's unique capabilities in high performance computing and AI.
• Reinforcement Learning: The reinforcement learning team focuses on model-based reinforcement learning with Bayesian methodologies, which are particularly apt for scientific settings, which demand utmost data efficiency, as experiments are necessarily expensive.  Reinforcement learning methods are used in many scientific applications, such as in materials synthesis to provide automated guidance of synthesis trajectories towards desired material properties.
• Scalability: The scalability efforts focus on developing novel scalable machine learning algorithms on leadership-class computing resources such as the world’s fastest supercomputer, Summit. A team of domain-specialists and computer scientists leverages the vast troves of data generated within the DOE complex and work together closely to push the performance boundaries of these high-performance machine learning methods by improving the training speed and inference quality in a variety of large-scale science and technology applications.