First of all, I should clarify that hydrogenolysis of epoxides does not always proceed with such predictable regioselectivity simply based on a switch in catalyst. Often it depends on the substituents present, as well as their relative configuration (for example, cis/trans for 2,3-disubstituted epoxides).[1]
In Reaxys I find no records of the epoxide in the question (2,2,3-trimethyloxirane) being hydrogenated. However, I did manage to find one study in which the authors measured the rates of hydrogenolysis of propylene oxide over Pt and Ni, finding that Pt gives more 2-propanol and Ni gives more 1-propanol.[2]

On top of that, there are also a few more examples that show similar reversals in regioselectivity for Pd (not Pt) and Ni.[3,4]

Bartók and Notheisz have put forward an explanation for the difference in regioselectivity between Ni and Pd/Pt.[2] They suggested that the key step is an oxidative addition of the metal into the C–O bond in the epoxide. This is fairly logical, considering the weakness of the C–O bond in a strained three-membered ring. For a Pt catalyst, the regioselectivity is determined by steric effects and insertion into the less sterically hindered C–O bond is preferred:

On the other hand, the authors write that for Ni
Because of the higher electron affinity of Ni, its insertion becomes ionic, formation of the surface species is controlled by electronic factors, and rupture of the C–O bond precedes formation of the Ni–C bond.
In modern parlance the way of expressing this is that the oxidative addition is asynchronous: the two new bonds are not being formed at exactly the same time. Because of the greater electron affinity of Ni compared to Pt, the Ni–O bond forms earlier than the Ni–C bond. If we were to draw a mechanism using curly arrows, it would be something like this: the flow of electrons depicted by the red arrow occurs earlier than the green arrow.

Consequently, there is a buildup of positive charge on the carbon undergoing bond breaking (hence "ionic"). This is not unlike the acid-catalysed opening of epoxides. The "electronic factors" that the authors refer to is simply the fact that the more substituted carbon is better able to carry this positive charge, which leads to the reverse in regioselectivity.
References
Murai, S.; Murai, T.; Kato, S. Reduction of Epoxides. In Comprehensive Organic Synthesis; Trost, B. M., Fleming, I., Eds.; Pergamon: Oxford, U.K., 1991; Vol. 8, pp 871–893. DOI: 10.1016/B978-0-08-052349-1.00249-3.
Bartók, M.; Notheisz, F. Stereochemistry of the hydrogenolysis of oxacycloalkanes on metal catalysts. J. Chem. Soc., Chem. Commun. 1980, 667–668. DOI: 10.1039/C39800000667.
Mitsui, S.; Kudo, Y.; Kobayashi, M. Stereospecific hydrogenolysis of benzyl-type alcohols. Tetrahedron 1969, 25 (9), 1921–1927. DOI: 10.1016/S0040-4020(01)82813-1.
Augustine, R. L. Organic Functional Group Hydrogenation. Catal. Rev.: Sci. Eng. 1976, 13 (1), 285–316. DOI: 10.1080/00087647608069940.