SciPy

NumPy provides the array; SciPy provides the algorithms. Submodules: optimize (minimization, curve fitting, root finding), signal (filters, convolution), stats (distributions, hypothesis tests), linalg (extended linear algebra), sparse (sparse matrices and graph algorithms), integrate, fft, spatial (KD-trees, convex hulls). Most of it is thin wrappers around battle-tested Fortran and C code — LAPACK, BLAS, ODEPACK, QUADPACK — which is exactly the point: fifty-year-old numerical routines with a decent API. Lends its name to the whole SciPy ecosystem including pandas, scikit-learn and Matplotlib.