This talk explores strategies to optimize tangential flow filtration (TFF)–based perfusion systems to support efficient continuous biomanufacturing. It will highlight approaches to improve process productivity and streamline operations while reducing equipment demands and facility footprint.

A 785 nm Raman-based model was developed as an online measurement for the detergent concentration in a continuous viral inactivation (VI) step. In the VI step, detergent is continuously added to a concentrated bioreactor perfusate stream. The Raman flow cell was first calibrated using a bank of previously generated perfusate samples spiked to varying detergent levels to generate a PLS model with an estimated accuracy of 0.02% w/v.

A digital shadow is a near–real-time, data-driven virtual representation of a bioreactor and its process. An AI-augmented digital shadow (AIS) accelerates process intensification by integrating multivariate data with machine-learning and hybrid mechanistic models to identify actionable parameters and nonlinear interactions. AIS enables what-if simulation of equipment configurations, CPP strategies, and feeding profiles before execution. It can infer hard-to-measure state variables, assess scale equivalence, and support model-predictive and adaptive control by recommending near-optimal setpoints. By mapping high-dimensional parameter space and ranking limiting phenomena, AIS strengthens design-space definition, regulatory justification, first-time-right success, and manufacturing robustness.