This presentation covers the development and scale-up of intensified fed-batch processes, highlighting key strategies for increasing productivity and process consistency across manufacturing scales.

Intensified fed-batch and high-density seed strategies are increasingly adopted to improve upstream efficiency while minimizing operational complexity. This presentation demonstrates how N-1 perfusion, smart media design, and PAT-enabled control can be combined to convert low-seed platform processes into intensified fed-batch processes. Across multiple programs, this strategy achieved ~2X improvements in space–time yield and ~30% shorter production durations, while maintaining product quality and avoiding full perfusion at production scale.

This presentation introduces mechanistic modeling frameworks developed to predict the complex dynamics of multispecific co-culture systems, providing a quantitative foundation for understanding and monitoring process behavior across varying conditions.

In Sanofi’s AAV manufacturing platform, the harvest process typically involves a flocculation step, where mixing plays a critical role. Developing a robust mixing scale‑down model is essential. Floc formation dynamics were characterized in real time using EasyViewer technology. The developed scale‑down model demonstrated excellent predictive performance. Implementation of the optimized flocculation conditions resulted in a two‑fold increase in depth filter throughput.

Digital twins are valuable for efficiently transferring complex processes from the laboratory to manufacturing scale and ensuring consistent results at different manufacturing sites. Here, we discuss case studies in the application of cell culture digital twins to tech transfer programs and demonstrate how modeling has allowed us to meet aggressive timelines and better serve the patients who need our products.

This talk will present the application of AI/ML to enhance mAb production in CHO cells. This AI tool automates rapid extraction of data from native file formats into structured templates powered by LLMs and performs in silico simulations to recommend optimal conditions for user-defined targets. By enabling intuitive and efficient exploration of complex datasets, the platform democratizes data access, accelerates insight generation, and supports data-driven decision-making.

This presentation explores how digital twins, combining mechanistic process understanding, AI, and process data, enable smarter, faster bioprocess development and control. Six industrial use cases demonstrate the value of digital twins: accelerated biosimilar development using PAT and glycan modeling; reduced experimental effort in viral vector process design; media optimization through time-resolved nutrient uptake prediction; UF/DF development guided by digital membrane and recovery modeling; scale-up informed by CFD-based reactor behavior; and fully integrated digital control of continuous bioprocesses sustained for over 30 days. Together, these examples show how digital twins streamline experimentation, enhance decision-making, and de-risk scale-up—unlocking end-to-end process insight from early development to production.

The future of bioprocessing is autonomous. I will show how a CHO digital twin fuses genome-scale metabolic modeling with PAT-driven AI to forecast VCD and titer in real time. By introducing an XAI-guided, BO–enabled adaptive control framework, we move to closed-loop decision-making by updating recipes and feeding towards desired setpoints trajectories. The result is interpretable, high-performance control that enables transparent, end-to-end bioprocess optimization.

• The Engine: Autonomous DT coupling with PAT and soft sensors
• The Intelligence: CHO GEM and AI forecasting for real-time cellular state prediction
• The Execution: An XAI-guided control framework bridging the gap between machine learning and operational trust