Many novel biologic-based therapeutics are progressing through the development process toward EMA and FDA filings and future approvals. The long-term viability of these new therapeutics partly depends upon the development of efficient and robust bio-manufacturing processes. Characterization, optimization, and long-term control of these processes is a requirement and development and then implementation of effective predictive models support this requirement. Traditionally, characterization is costly, time consuming, and often no effective predictive models result from these efforts. We propose by using highly efficient experimental designs and predictive modeling methods from machine learning, cost effective and reliable predictive models can be developed. The models subsequently may be used for scale-down modeling, process characterization, optimization, and identification of critical process parameters for manufacturing control strategies. Using highly efficient space filling experimental designs combined with a novel method of predictive model developed known as self-validating ensemble modeling (SVEM) combined with artificial neural networks (ANNs) allows for predictive models to be designed that are more representative i.e. predictive.