The Gaussian Process (GP) is arguably the leading uncertainty-quantification tool at the disposal of machine learning engineers and spatial statisticians. Its one major disadvantage is scalability which limits competitiveness with other machine learning methodologies. The Python library gp2Scale — part of the fvgp package (pip install fvgp) — makes exact GPs scalable to millions of data points via kernels that let the algorithm discover naturally occurring sparsity in datasets.