Creating an AI Solution for Certificate Verification Using Amazon Bedrock with Tenthpin

As an SAP Consulting Service Partner, Tenthpin offers SAP consulting and rollouts along with data and analytics, quality assessment and assurance, and other solutions and services to large and midsize life sciences companies around the globe. Tenthpin’s solutions include Tenthpin Intelligent Quality Suite, a set of AI-based tools for quality management and certificate generation that helps customers more efficiently meet regulatory requirements. Tenthpin identified another opportunity to meet its customers’ needs: comparing specification documents with certificates generated by suppliers and contract development and manufacturing organizations (CDMOs).

In manufacturing a drug, a pharmaceutical company lists the drug’s composition, appearance, and other information in a specification document. A CDMO then produces the drug according to the specifications and creates a detailed certificate for each batch produced. After receiving a batch, the pharma company will compare the CDMO’s certificate with the specification document. This is also relevant for the procurement of any other materials. Thus, for every batch of drugs, a pharma company needs to make sure that the details in both documents match and determine whether any deviations—such as a slightly different color—are acceptable before approving the certificate from the CDMO. “To avoid mistakes, some companies even apply the four-eye principle, which requires that two people do the verification and come to the same result,” says Thomas Weber, chief product officer at Tenthpin Solutions, which is the product arm of Tenthpin that creates software solutions for life sciences companies.

Big pharma companies need to verify more than 100,000 certificates each year. So, Tenthpin Solutions decided to build a generative AI solution that would enhance accuracy and speed up this critical process and started comparing different solution providers. Having previous experience with AWS services, the company chose AWS because of its stability, reliability, and transparent cost structure. The company also wanted to use the performant and cost-effective infrastructure for generative AI that AWS provides.

Tenthpin Solutions implemented Amazon Bedrock—the simplest way to build and scale generative AI applications with foundation models—for T/ICV. The solution also uses AWS Lambda—which lets businesses run code without thinking about servers or clusters—for workflow, automation, and integration tasks.

The company used its own software development life cycle and quality management system to meet regulatory requirements, such as GxP—a set of good practice guidelines—as it built and implemented the architecture. After testing the solution, Tenthpin released the first GxP-ready version after 6 months. “The level of technical understanding in combination with business understanding is remarkable at AWS, which is very customer oriented,” says Weber.

T/ICV is a software-as-a-service solution that is provided through SAP Business Technology Platform. Tenthpin’s customers have single sign-on access to the platform, to which they can send specifications and certificates either by manual upload or by automatic syncing from data storage or an email mailbox. T/ICV then suggests matching specification documents, and users can upload their own specification documents, too. T/ICV extracts text from the certificate and the specification document to perform a comparison, creates a workflow instance, and notifies the user when the comparison is complete.

The company stores the data that T/ICV gathers from the documents in Amazon Aurora, a relational database management system built for the cloud. Based on this data, T/ICV can run the required comparisons and decide whether to give approval. “T/ICV supports the user in making the approval decision,” says Weber. “It is not fully automated because we operate in a regulated environment.”

T/ICV reduces the time needed to verify certificates by 95 percent and increases the accuracy of comparison to nearly 100 percent for many certificates. It has built-in regulatory compliance features, such as audit logs, as well. “Customers say that T/ICV is simple to use and reliable because it generates a workflow task that must be completed to move to the next stage,” says Weber. “It has increased quality and reduced risk for our customers.”

The company uses a multitenancy environment on AWS to accelerate customer onboarding. T/ICV already meets regulatory requirements, and customers can go live within 2 weeks of purchase and environment provision, instead of several months. They can scale usage according to their needs and pay only for the certificates they process through T/ICV.

Having added T/ICV to the Tenthpin Intelligent Quality Suite, Tenthpin Solutions plans to develop new solutions to meet customer needs and expand into other markets. It is cooperating on a second release of T/ICV with a major life sciences company to incorporate the verification of sterilization certificates, which are issued when companies need to sterilize medicinal products. Tenthpin Solutions expects to release the new version within 4 months of the first release.

An upcoming release will connect data and analytics to provide insights and sophisticated reporting. “The certificates in T/ICV have given us access to data points in a structured format, which our customers didn’t have earlier,” says Weber. “This is a game changer that we didn’t consider when we began, but our release will focus on using this data and connecting it to additional systems.”

By innovating on AWS, Tenthpin Solutions has set itself up for continued improvement. “Collaborating closely with the AWS team, we can get the latest information, choose the right technologies, and tap into expert guidance,” says Weber. “With support from AWS, we’ve increased the development speed and stability of T/ICV.”