A leading pharmaceutical company streamlines its production processes

The pharmaceutical manufacturing process is incredibly complex. A single facility may be required to manufacture medicines in a number of different formats, from blister packs to vials. Different batches of drugs are often interdependent, so a facility cannot begin production on one batch before completing another. And manufacturing each batch involves multiple substeps, each of which introduces its own complexities. Precise planning is paramount in order to deliver life-saving products efficiently and on schedule.

For this project, the pharmaceutical company wanted to use AI and machine learning to improve planning and enable its production facilities to run more efficiently, with fewer delays. Specifically, the company wanted AI models to predict the duration of batches and other steps of the manufacturing process — including wait times between steps — more accurately than the original planned timelines. This would enable human planners to make more accurate decisions about when to schedule production for particular batches.

To achieve this goal, the pharmaceutical company needed AI and machine learning expertise. It also needed to gather data from multiple systems to build a robust data pipeline that would deliver clean, high-quality data to be ingested by the models.

The project leveraged Orbis's data science expertise to build AI models that provide insights into the pharmaceutical company's manufacturing process.

The team’s goal was to build four models, each of which would “zoom in” further to examine manufacturing processes with a greater degree of granularity. Each model would predict duration at different granularities, as well as the waiting time between different batches. The team is continuing to work to deploy further models as more data becomes available.

Today, the initial models have been implemented at both production sites and are already helping deliver efficiency gains. A dedicated user interface (UI) enables human planners to review the models’ predicted durations for different parts of the process.

These highly accurate predictions let planners optimize their work and avoid unexpected delays or interdependent batches falling out of sync, ensuring that drugs reach pharmacies, hospitals and — most importantly — patients on time.

This project demonstrates the power of AI and machine learning across the pharmaceutical value chain. Pharmaceutical companies have a major opportunity to unlock efficiency and gain new insights through the use of technology everywhere from R&D labs to production lines.

The collaboration with the pharmaceutical company is continuing to grow. The next step is to adapt the existing models to flag delays and potential issues in the drug production process as well as predicting process duration. Eventually, the company hopes to expand the use of these models to additional sites, multiplying the impact of the project for the long term.