Demo code on how to use FFT to sinc-interpolate (upsample). (You probably should just use scipy.signal.resample which will filter and window the signal to avoid the ringing that this FFT-only method invariably yields, unless you know what you're doing.)

fftUpsample.py

import numpy as np
from numpy.fft import fft, ifft


def fftUpsample(a, factor):
    a_f = fft(a)
    n = len(a)
    assert n % 2 == 0, "odd case is harder"
    padded = np.zeros(int(len(a) * factor), dtype=a_f.dtype)
    padded[:n // 2 + 1] = a_f[:n // 2 + 1]
    remaining = n - (n // 2 + 1)
    padded[-remaining:] = a_f[n // 2 + 1:]

    padded[n // 2] /= 2
    padded[-remaining - 1] = padded[n // 2]

    return factor * ifft(padded)