Magnetic fields must be known for describing many astrophysical processes, including star formation. The knowledge of magnetic fields is also essential for understanding the properties of the foreground galactic radiation interfering with the Cosmic Microwave Background (CMB) polarized signal. I shall discuss a new way to approach this problem using the innovative technique of velocity gradients. The latter technique employs the properties of magneto hydro dynamic (MHD) turbulence to map the magnetic fields using spectroscopic data. I shall briefly discuss how Machine Learning can improve the ability of the technique to separate the contributions from turbulent regions and regions undergoing the gravitational collapse.