Graduate students (master’s and PhD) in any STEM field can learn applied AI for their research in Spring 2027 in PHY 591 AAI4STEM: Applications of AI Methods Across STEM Disciplines. To enroll, register for PHY 591-0001, listed in the catalog as “Special Problems.” The course meets Tue & Thu 11:00–12:15 in Swan Hall 211 and is capped at 20 students.
What you need: linear algebra, multivariable calculus, basic probability and working Python (NumPy, Matplotlib, Jupyter). No prior deep learning is required; the course opens with four weeks of foundations. Advanced undergraduates may enroll by permission. Taught by Michael Pürrer (Physics), 3 credits, January 26 – May 3, 2027.
What you’ll do: AlphaFold, AI weather forecasting and gravitational-wave inference all rest on a small set of ideas. In this course you build those ideas from scratch in PyTorch. You then test them on real scientific data against the classical methods the AI is supposed to beat. There are seven method modules, each pairing an architecture with a science:
- Graph networks for molecules
- Physics-informed networks for coastal waves
- Neural ODEs for measles epidemics
- CNNs and transformers for genomics
- Normalizing flows for black-hole ringdowns
- Diffusion models for gap-filling ocean satellite data
- Reinforcement learning for building energy
Most of the work goes into a team project that grows into a capstone presented in the final week. Projects run on laptops first and move to URI’s UNITY cluster when needed.
Questions about whether the course is a good fit are welcome: mpuerrer@uri.edu. See the attached flyer for details.
