A major challenge to bringing artificial intelligence (AI) into bioprocessing is a need for cultural change within companies. That’s according to Varsha Daswani, PhD, a senior director at data science consultancy, Lumilytics. Understanding how AI and data science can help with biomanufacturing is essential to reaping its benefits, she says. And success often depends on unglamorous data preparation.
“The gap we’re trying to bridge is between bioprocessing specialists and experts in data science,” she explains. “We need [to] communicate across that gap and find ways to integrate data science into workflow.”
Daswani, who will be speaking at several sessions at the forthcoming Bioprocessing Summit in Boston, reports a huge variation in clients’ knowledge about what artificial intelligence can do.
“We get the whole spectrum from people who want to implement process modeling or a digital twin, to people who want ChatGPT to run a bioreactor,” she continues. “Both are okay. We can come in at any point, meet people where they are, and—even if they have no knowledge of AI—we love enthusiasm.”
Standard format for data
Critical to making projects a success, she says, is doing the unfashionable work of ensuring data is in a standard format and clearly defining how it will be used.
“Capabilities are also important,” stresses Daswani. “You need to establish [what you want] the technology to do. Many people we see are unfamiliar with how the technology works.”
Establishing a use case can be about assessing a bioprocess to find bottlenecks and then determining which technology and tools can help. According to Daswani, Lumilytics’ clients typically work on large molecules or antibodies, although the company is flexible. She’s also keen to emphasize that Lumilytics can help with end-to-end process optimization, as well as smaller-scale projects such as automating chromatography data analysis.
“Our favorite and most successful approach is to see what the problem is and build a customized solution in-house,” she says. “In the past, companies have offered out-of-the-box solutions where they show people what to do and then they leave but this often leads to the customer struggling with implementation.”
