In this second of the two-part paper on the development of Parametrically Homogenized Constitutive Models (PHCMs), constitutive coefficients are represented in functional forms of the representative aggregate microstructural parameters (RAMPs), e.g. texture intensity parameters, misorientation parameter and grain size, identified in the first part (Kotha et al., 2019). Efficient and accurate PHCMs are developed for finite deformation, anisotropic thermo-elasto-plastic behavior of polycrystalline Ti6242S alloys. The constitutive framework in the PHCM is chosen to reflect fundamental characteristics of elasto-plastic anisotropy, tension-compression asymmetry, anisotropic hardening and rate-dependent mechanical behavior, observed in Ti alloys. The functional forms of the PHCM constitutive coefficients in terms of the RAMPs are deciphered using a machine learning code Eureqa. The machine learning data-base is generated through detailed micro-mechanical crystal plasticity FE simulations of microstructure-based statistically equivalent RVEs or M-SERVEs of the polycrystalline material. A variety of M-SERVEs with different morphology and crystallography characteristics and thermo-mechanical loadings are considered. The proposed PHCM is numerically implemented in ABAQUS as a user subroutine. The accuracy of the PHCM model is demonstrated by validating the PHCM predictions for Ti6242S with tensile test experiments at different strain-rates and temperatures.