Drug Production Process Optimization

Modern bioreactors are often equipped with a suite of sensors which gather information such as nutrient concentration, product yield or waste product accumulation. The optimal parameters can vary between individual reactors, batches and products. Depending on the monitoring frequency the amount of generated data can be very large. Additionally, non-linear relationships between key variables can be difficult to interpret for production engineers. As machine learning talent is a scarce resource, especially for small to mid-size companies, production engineers are frustrated by the untapped potential of their data.

Automated machine learning allows production engineers and other domain experts to regularly train machine learning models. Predictions generated by these models allow for timely intervention, optimal set up, and time-dependent forecast of product yields.

Modulos AutoML automatically selects the appropriate models, feature engineering algorithms and tunes the hyper-parameter. All models are openly accessible to the user and can be deployed instanteanously. With the intutitive and easy to use user interface the process engineers do not have to waste their time on manually coding high performing models.