Explore the fusion of traditional mathematical modeling with machine learning in scientific machine learning (SciML) through this conference talk. Delve into the use of neural network function approximations to bridge the gap between ordinary differential equations (ODE) compartmental models and epidemiological agent-based models (ABM). Learn how universal differential equations (UDE) surrogates preserve global disease dynamics while isolating local behaviors of ABMs. Discover the process of automating hierarchical calibration of ABMs, separating global parameter estimations from local parameter influence. Gain insights into the emerging field of SciML and its applications in handling challenges such as numerical solver implementation, model-form error estimations, and computational expense reduction in high-fidelity models.