Steel production accounts for approximately 20–25% of global industrial CO₂ emissions [1], making it one of the most carbon-intensive industrial sectors. Accurate and efficient prediction of steel creep models is therefore crucial for optimizing material performance, extending service life, and supporting sustainability-driven engineering design. However, model calibration remains challenging due to strong parameter coupling, nonlinear constitutive behavior, and experimental uncertainties, while conventional deterministic optimization methods often yield non-unique solutions and fail to quantify uncertainty. To address these limitations, we develop a Bayesian calibration framework implemented in our open-source Python library, ACBICI (A Configurable BayesIan Calibration and Inference package) [2]. Inspired by the Bayesian inverse modeling principles of Kennedy and O’Hagan [3], the framework provides a flexible inference environment that can incorporate numerical models of varying complexity directly into the calibration process. Model parameters are sampled dynamically, and corresponding simulations are performed within an adaptive inference loop. Relevant response features are extracted from the simulation outputs and used to update posterior distributions, enabling robust parameter calibration and providing insight into the underlying model behavior. The proposed framework shows promising predictive capability for steel creep behavior under uncertainty. By estimating posterior distributions of model parameters and hyperparameters that are consistent with experimental observations and numerical simulations, the methodology facilitates meaningful uncertainty quantification and may improve predictive reliability.