We propose a novel technique to separate the late-time, post-reionization component of the kinetic Sunyaev-Zeldovich (kSZ) effect from the contribution to it from a (poorly understood and probably patchy) reionization history. The kSZ effect is one of the most promising probe of the {em missing baryons} in the Universe. We study the possibility of reconstructing it in three dimensions (3D), using future spectroscopic surveys such as the Euclid survey. By reconstructing a 3D template from galaxy density and peculiar velocity fields from spectroscopic surveys we cross-correlate the estimator against CMB maps. The resulting cross-correlation can help us to map out the kSZ contribution to CMB in 3D as a function of redshift thereby extending previous results which use tomographic reconstruction. This allows the separation of the late time effect from the contribution owing to reionization. By construction, it avoids contamination from foregrounds, primary CMB, tSZ effect as well as from star forming galaxies. Due to a high number density of galaxies the signal-to-noise (S/N) for such cross-correlational studies are higher, compared to the studies involving CMB power spectrum analysis. Using a spherical Bessel-Fourier (sFB) transform we introduce a pair of 3D power-spectra: ${cal C}^{parallel}_ell(k)$ and ${cal C}^{perp}_ell(k)$ that can be used for this purpose. We find that in a future spectroscopic survey with near all-sky coverage and a survey depth of $zapprox 1$, reconstruction of ${cal C}^{perp}_ell(k)$ can be achieved in a few radial wave bands $kapprox(0.01-0.5 h^{-1}rm Mpc)$ with a S/N of upto ${cal O}(10)$ for angular harmonics in the range $ell=(200-2000)$ (abrdiged).