PCA projects data onto principal components to reduce dimensionality while retaining most information. It works by (1) zero-centering the data, (2) calculating the covariance matrix to measure joint variability, (3) computing eigenvalues and eigenvectors of the covariance matrix to identify principal components with most variation, and (4) mapping the zero-centered data to a new space using the eigenvectors. This transforms the data onto a new set of orthogonal axes oriented in the directions of maximum variance.