Polymer nanocomposites have drawn considerable attention due to the improvement of structural and functional properties of polymer based systems. The common way to elaborate nanocomposites is the “top-down” approach, i.e. the dispersion of fillers into the polymer. However, the agglomeration of nanofillers in the polymer and the increasing concerns about the manipulation of nanoobjects are the main issues of this technique. To overcome these problems, a “bottom-up” approach has been developed in recent years combining reactive extrusion and sol-gel chemistry. This approach has some limitations for non-polar polymers as polyolefins. Indeed, water is required and the surface of such inorganic phases presents hydroxyl groups involving a weak interaction with the polymer matrix. Our approach is to use non-hydrolytic sol-gel (NHSG) chemistry to generate nanoparticles by reactive extrusion. The specificity of NHSG is the use of an organic oxygen donor instead of water and the ability to control the shape and the size of nanoparticles. First experiments were carried out in apolar liquid medium as a model in order to evaluate the kinetics and to select the best candidates to the transfer to reactive extrusion. Based on these results, syntheses of TiO2 by the anhydride route were then carried out in molten polypropylene by reactive extrusion. Different oxygen donors and titanium precursors were investigated in order to improve the compatibility between matrix and fillers and obtain homogeneous dispersion of small size nanoparticles. The impact of the nature of the polymer matrix was also scrutinize and well as the possibility depending of the inorganic precursors to modulate the morphology on the inorganic object.