Polymer pyrolysis is vital in modern aerospace and other mechanical engineering systems, with char yield being pivotal to material performance. For example, Thermal Protection Systems (TPS) used on spacecraft during atmospheric reentry dissipate excess heat through the burning of polymer layers into char. Conversely, plastic gasification harnesses volatile gas products from polymer pyrolysis for power generation. Thus, char yield is essential to de novo material discovery for application-specific polymer use; however, experimental data on char yield is extremely limited by high costs and time constraints of high-temperature testing. In order to automate the discovery of different polymer char yields, the Sparks lab leverages Molecular Dynamics (MD) and Density Functional Theory (DFT) simulations. Their data independence and reliance on first-principle physics offers a promising pathway towards predicting pyrolysis characteristics, but holds expensive computational cost. Meanwhile, machine learning (ML) models offer low computational costs but require a massive database for training. In order to balance both the computational efficiency of MD/DFT simulations with predictive ML results, we present an integrated framework. By utilizing simulation-generated data to train machine learning models, this framework promises to streamline polymer pyrolysis characterization while bypassing traditional physical experiments.