The phaseout of high global warming potential refrigerants has environmental, economic, and engineering implications. These refrigerants are often azeotropic mixtures of high value hydrofluorocarbons (HFCs) which must be separated for their sustainable reuse and recycle. However, energy efficient and economically feasible azeotropic separations are a significant challenge in the chemical engineering community, often requiring the design of novel solvent materials, e.g., ionic liquids. Here, we propose a multi-scale design and optimization, ``data-to-design", framework for identifying an ideal IL separating agent for each HFC refrigerant mixture. This framework is an iterative loop in which data is generated and used to fit thermodynamic models and then perform preliminary process design calculations and technoeconomic analyses under uncertainty. A feedback loop guides data generation efforts towards valuable experiments and the thermophysical properties of optimal ILs.This data-to-design framework is an example of highly collaborative, integrated research that bridges the perspectives and tools of the experimental, molecular science and engineering, and process systems engineering communities to develop novel solutions to an ongoing chemical engineering challenge. Here, we highlight two examples of successful synergistic research efforts in the data-to-design framework. First, we demonstrate two workflows which harness machine learning and Bayesian optimization techniques to fit cheap-to-evaluate surrogate models as replacements for computationally expensive molecular simulations, enabling rapid calibration of molecular models for HFCs. Then, we propose a rigorous thermodynamic model selection and analysis workflow which utilizes data science tools, including information criteria, Fischer information matrix-based identifiability analyses, uncertainty quantification, and model-based design of experiments. This workflow enables the systematic selection of a thermodynamic model that is the most accurate, predictive, and interpretable from a library of candidate models for an HFC/IL system. Additionally, it guides data generation collaborators towards the most valuable measurements, facilitating more efficient allocation of experimental resources. Finally, we conclude with a prospective for applying the data-to-design framework to HFC/IL systems, but emphasize this framework and the tools presented here are applicable to many systems of interest in the chemical engineering community.