The injection of a spin-polarized current into a semiconductor, one of the key requirements for spintronics, poses the challenge for computational materials science to possibly screen suitable materials. In a number of theoretical investigations, we have put forward magnetic intermetallic compounds grown epitaxially on Si as promising candidates. We employed density functional theory calculations with the GGA-PBE exchange-correlation functional and the full-potential augmented plane wave plus local orbital (FP-APW+lo) method, as implemented in the WIEN2k package. In the spirit of computational materials science, we investigated the stability and magnetic properties of thin films of the Heusler alloy Co2MnSi, as well as of binary late transition metal monosilicides, in contact with the Si surface. For the Heusler alloy Co2MnSi, we could show that the (001) surface retains the half-metallic character of the bulk if a fully Mn-terminated surface is prepared. At interfaces with Si, a finite density of states at the Fermi energy was found for both spin channels, but the half-metallic behavior recovers only a few layers away from the interface. For the monosilicides of the late 3d-transition metals (Mn, Fe, Co, Ni), we predict a CsCl-like structure that has not yet been observed as bulk compound but may be stabilized epitaxially on Si(001). For very thin films of CoSi and MnSi grown in this structure, our calculations find a ferromagnetic ground state. Recently, we identified the atomic structure of MnSi films on Si(111) which is close to the natural crystal structure of bulk MnSi (B20), and also shows large magnetic moments of the Mn atoms at the surface and interface. All MnSi films have a high degre- e of spin polarization (between 30% and 50%, depending on film thickness) at the Fermi level, and are thus promising materials for fabricating electrical contacts for spin injection into Si.