AARDVARK: Adaptive Automation and Real-time Data Visualization for ARPES Research Kit

Angle resolved photoemission spectroscopy (ARPES) is an information-rich and multi-dimensional experimental technique where AI/ML is poised to enhance the efficiency of data collection and analysis. Typical ARPES experiments begin with a time-intensive procedure to manually search for the most spectroscopically 'ideal' region of the sample. Our work creates a modular AI/ML pipeline that autonomously controls the experiment during this initial search. This pipeline includes modular steps for data dimensionality reduction, clustering, next-measurement prediction, denoising, and visualization. This speedup increases productivity of users by giving more time for important measurements and less time on manual searching. We will discuss our pipeline and examples of its applications to specimens with physically relevant spatial inhomogeneity and comparisons to traditional techniques.