A specialist biologics manufacturer is planning to use generative AI and natural language processing (NLP) throughout their organization to help staff work more efficiently.
ADMA Biologics, whose director Robert Brooks gave a talk at BioProcess International last month, is currently rolling out machine learning software originally designed to optimize production processes.
According to Brooks, ADMAlytics™ was initially intended to automate and standardize combining plasma from multiple donors—previously a highly manual process.
“ADMA identified various functions that we believed could be enhanced with machine learning,” he writes, in a response to GEN.“These include using ADMAlytics to optimize the plasma pooling process.”
Initial uses, include helping ADMA to adjust pooling strategies to account for changes in plasma supply, demand, and inventory. However, he explains, the company swiftly realized the software had further uses in predicting the product yields from each batch of plasma.
“We quickly identified expanded use cases which use historical data to predict yields of future batches based on the proposed pools from ADMAlytics,” he points out. “This allows [us] to optimize production yields and reduce waste, therefore leading to better planning for raw material procurement, improved operational efficiency as well as cost savings and improved sales forecasting.”
The company is now looking to use the software, which is based on tcgmcube™, a generative AI for handling and analyzing complex data, to work more efficiently right across their commercial operations.
“The rise of generative AI and natural language processing presents a significant opportunity to embed these technologies in ADMA employees’ daily work, enabling them to focus on higher-value tasks and work more efficiently,” according to Brooks, who adds that the company has also used generative AI to deploy a chatbot. This relies on NLP to understand search queries based on the company’s electronic quality management system to provide more precise and relevant results.