Measuring the relative mass contributions of luminous and dark matter in spiral galaxies is important for understanding their formation and evolution. The combination of a galaxy rotation curve and strong lensing is a powerful way to break the disk-halo degeneracy that is inherent in each of the methods individually. We present an analysis of the 10-image radio spiral lens B1933+503 at z_l=0.755, incorporating (1) new global VLBI observations, (2) new adaptive-optics assisted K-band imaging, (3) new spectroscopic observations for the lens galaxy rotation curve and the source redshift. We construct a three-dimensionally axisymmetric mass distribution with 3 components: an exponential profile for the disk, a point mass for the bulge, and an NFW profile for the halo. The mass model is simultaneously fitted to the kinematics and the lensing data. The NFW halo needs to be oblate with a flattening of a/c=0.33^{+0.07}_{-0.05} to be consistent with the radio data. This suggests that baryons are effective at making the halos oblate near the center. The lensing and kinematics analysis probe the inner ~10 kpc of the galaxy, and we obtain a lower limit on the halo scale radius of 16 kpc (95% CI). The dark matter mass fraction inside a sphere with a radius of 2.2 disk scale lengths is f_{DM,2.2}=0.43^{+0.10}_{-0.09}. The contribution of the disk to the total circular velocity at 2.2 disk scale lengths is 0.76^{+0.05}_{-0.06}, suggesting that the disk is marginally submaximal. The stellar mass of the disk from our modeling is log_{10}(M_{*}/M_{sun}) = 11.06^{+0.09}_{-0.11} assuming that the cold gas contributes ~20% to the total disk mass. In comparison to the stellar masses estimated from stellar population synthesis models, the stellar initial mass function of Chabrier is preferred to that of Salpeter by a probability factor of 7.2.