We are presenting the first ab initio structure investigation of the loosely bound 11-Be nucleus, together with a study of the lighter isotope 9-Be. The nuclear structure of these isotopes is particularly interesting due to the appearance of a parity-inverted ground state in 11-Be. Our study is performed in the framework of the ab initio no-core shell model. Results obtained using four different, high-precision two-nucleon interactions, in model spaces up to 9\hbar\Omega, are shown. For both nuclei, and all potentials, we reach convergence in the level ordering of positive- and negative-parity spectra separately. Concerning their relative position, the positive-parity states are always too high in excitation energy, but a fast drop with respect to the negative-parity spectrum is observed when the model space is increased. This behavior is most dramatic for 11-Be. In the largest model space we were able to reach, the 1/2+ level has dropped down to become either the first or the second excited state, depending on which interaction we use. We also observe a contrasting behavior in the convergence patterns for different two-nucleon potentials, and argue that a three-nucleon interaction is needed to explain the parity inversion. Furthermore, large-basis calculations of 13-C and 11-B are performed. This allows us to study the systematics of the position of the first unnatural-parity state in the N=7 isotone and the A=11 isobar. The 11-B run in the 9\hbar\Omega model space involves a matrix with dimension exceeding 1.1 x 10^9, and is our largest calculation so far. We present results on binding energies, excitation spectra, level configurations, radii, electromagnetic observables, and 10-Be+n overlap functions.