Lubricated gears and bearings operate under the Thermo-Elastohydrodynamic lubrication (TEHL) regime, characterized by, an extremely thin liquid oil film of 50nm-1μm thickness, separating the solid surfaces. In this film, the hydrodynamic pressure may rise up to 4GPa, whereas shear rates may easily exceed 109s-1 and temperatures rise up to 250°C. Under these extreme conditions, the lubricant properties strongly diverge from those at atmospheric conditions. Both mechanical properties (viscosity, density), as well as thermal properties (thermal conductivity, heat capacity) display non-linear dependency on the pressure, temperature and shear rate. Although those thermomechanical properties are decisive for the film thickness, lubrication efficiency, and component lifetime, it is very difficult to measure them experimentally. Hence the exact constitutive behavior remains mostly a blind spot for lubricant suppliers, gearboxes and bearing OEMs, machine builders, and tribologists. The objective of the current proposal is to integrate the strength of combined first-principles atomistic and molecular modeling with machine learning to acquire the correct thermomechanical properties of Polyalphaolefins, a modern synthetic lubricant, under relevant TEHL conditions.