Building Scalable AI for CMC: QbDVision + Amazon Bedrock

QbDVision uses Amazon Web Services (AWS) and Amazon Bedrock to help pharmaceutical organizations transform fragmented Chemistry, Manufacturing, and Controls (CMC)  information into a connected Digital CMC knowledge foundation. Built on structured, governed data, the platform enables scalable AI for GxP environments that helps scientists and technical teams find answers faster, preserve organizational knowledge, improve collaboration, and make more confident development decisions.

 

Many great innovations begin with solving a real-world problem. When Yash Sabharwal (CEO) and Ryan Shillington (CTO) founded QbDVision in 2017, they set out to solve one of pharmaceutical development's biggest challenges: critical Chemistry, Manufacturing, and Controls (CMC) knowledge trapped within disconnected documents, spreadsheets, and data systems. They believed CMC shouldn't be the bottleneck between scientific discovery and patient delivery, and that better structured knowledge could help pharmaceutical organizations bring new therapies to market faster.
At the time, Sabharwal was leading the operations of a complex combination product development program, with ingredients sourced in Switzerland, formulation manufacturing in California, auto-injector assembly in Florida, and packaging in Michigan.

Coordinating work across these regions highlighted a fundamental challenge. Every stage generated valuable scientific and manufacturing knowledge, yet that knowledge remained fragmented across teams, documents, and disconnected systems. Instead of building on what was already known, experts spent valuable time searching for information, recreating work, validating data, and transferring knowledge between functions. The result was slower decision-making, duplicated effort, and longer development timelines.

“I didn’t have visibility across the entire program,” says Sabharwal. “Managing CMC processes was inefficient and time-consuming. I remember thinking, ‘Does every company waste time and effort like this?’”

Coming to the pharmaceutical industry as an outsider, Shillington saw the same challenge from a different perspective. “The sheer number of people required to get a drug to market is staggering,” he says. “Thousands of specialists, often working in separate systems, with separate documents, and different business cultures.”

Together, they realized the industry wasn't lacking scientific expertise. It was struggling to connect and reuse it. That insight became the foundation for QbDVision's Digital CMC platform: transforming document-based, fragmented data management into structured, connected knowledge using network models that better represent complex systems like pharmaceutical manufacturing. This approach  enables cross-functional collaboration and provides a trusted foundation for the next generation of AI-powered drug development.

These challenges became the starting point for QbDVision’s Digital CMC concept: a platform designed to manage multiple dimensions of information while also capturing the evolution of this information over the lifecycle of development. By shifting from a documents/data mindset to a network/knowledge mindset, it forces a shift from thinking about requirements individually to thinking about them systematically.  

“What I wanted to build was the platform I needed in the industry,” says Sabharwal. “Not just somewhere to store information, but a way to make it accessible and useful.”

From the start, the team knew two things were critical. Drug-development data can be worth billions, so security had to be built into the platform from the beginning. And, given the enormous volumes of unstructured information in the sector, QbDVision had to be able to scale as its customers and their programs grew.

For Shillington, AWS was the obvious choice. “I’ve been using AWS for a very long time and I’m a huge fan,” he explains. “If you’re starting a new company from scratch, you want to be on AWS. Its services allow you to scale the platform, support growing data volumes, and meet enterprise security requirements without adding unnecessary infrastructure overhead.”

QbDVision built its platform on AWS and now uses a portfolio of cloud services, including Amazon CloudFront, AWS CloudFormation, Amazon CloudWatch, AWS API Gateway, AWS Lambda, AWS RDS, and Amazon S3 to support security, scalability, and modern development practices. As the business has grown, AWS has helped the team rely on managed and serverless services rather than spending time and resources managing infrastructure.

One service Shillington highlights is Amazon GuardDuty, AWS’s threat detection service, which helps identify suspicious activity and potential security risks.

“GuardDuty is really exceptional from a security perspective,” he says. “With our focus on large enterprise customers, GuardDuty helps us meet and exceed their security requirements.”

 

QbDVision has become a market leader by developing a fit-for-purpose knowledge and workflow management platform specifically for the industry. Moving to the network models concept where information in documents is deconstructed into nodes that are connected with predefined relationships is a powerful construct.  The idea of the knowledge graph is built on this inherently extensible and scalable concept. The network model is also well suited to deploying AI, which is where QbDVision is going next.   

“The initial value of QbDVision starts with its role as a CMC system of record,” Sabharwal explains. From this starting point, QbDVision’s value grows as a system of work enabling key digital workflows. “As knowledge and capabilities multiply, QbDVision creates a strong foundation of high quality structured and contextualized data with digitalized business processes, the two key ingredients for successful AI transformation.” In other words, QbDVision evolves from a system of record to a system of work to a system of insights.

This evolution to a system of insights is already underway as QbDVision is launching AI-based tools sitting on top of its network/workflow model to convert unstructured to structured data at scale which are then mined for knowledge as opposed to simple data searching. Instead of relying on disconnected documents, the platform creates a connected knowledge foundation that enables AI to understand relationships across products, processes, procedures, materials, and manufacturing. This allows teams to retrieve trusted insights, accelerate decision-making, preserve organizational knowledge, and apply intelligence throughout the drug development lifecycle.

In life sciences, incorrect data can have life-and-death consequences. That’s why AI use can only be contemplated when it can be trusted to produce reliably correct answers, and all AI outputs are checked and verified. “Getting it right is really important; our customers won’t use it if they can’t trust it,” says Shillington. Sabharwal puts it simply. “With AI, it’s garbage in, garbage out. You can’t change that fundamental aspect of processing data.”

This is why QbDVision does not simply place AI on top of scattered PDFs, spreadsheets, and manual workflows. It structures CMC data first, so AI tools can work with cleaner, more meaningful information.

“It’s about finding genuine answers,” says Shillington. “If you’re working with a traditional manual CMC workflow and you need an answer to an important question, the answer you get will, at best, be an approximation. You’ll need to do the legwork to get the answer. But with QbDVision’s Qurio, a suite of AI-enhanced Digital CMC tools, you can get the answer almost instantly, with a very high degree of accuracy.”

That distinction is critical in regulated CMC. The goal is not to present AI as infallible, but to ensure its outputs are credible for their intended use where credibility is based on accuracy and precision.  By combining structured CMC knowledge, credibility assessment, governance, and human oversight, QbDVision helps organizations adopt AI in a way that is measurable, reviewable, and aligned with the expectations of regulated drug development.

QbDVision builds its AI capabilities on Amazon Bedrock, giving the team access to a broad ecosystem of leading foundation models while maintaining the flexibility to evolve as model performance improves. "We started our process by designing for accuracy and precision from the beginning. Amazon Bedrock gives us the flexibility to evaluate and use the models best suited for each capability without being tied to a single provider," says Shillington. "Combined with AWS's security and governance capabilities, it provides the trusted infrastructure we need to deliver scalable AI for regulated CMC environments."

Amazon Bedrock also supports the company’s need for secure, measurable AI in regulated workflows. QbDVision uses Bedrock as part of its AI-based import process, helping turn information trapped in PDFs into structured, auditable CMC records. Bedrock’s data isolation is also critical: prompts, documents, and outputs remain within the AWS environment and are not used to train underlying models.

Just as importantly, AWS has helped the team rethink how it evaluates AI outputs, moving from generic classification scores toward measures such as completeness, correctness, and traceability using Bedrock Evaluations. This is especially important in GMP workflows, where quality, safety, and traceability are essential.

Looking to the future, QbDVision sees a world where AI helps pharmaceutical customers solve more workflow challenges themselves, with Digital CMC providing the trusted foundation those workflows need. 

Shillington believes AI can transform complex CMC workflows, but not in isolation. Organizations first need a structured, governed system of record that connects their data, processes, and scientific context. “AI alone cannot solve the foundational challenges of CMC,” he says. “The breakthrough comes from combining AI with structured, contextualized knowledge. That is what enables organizations to generate credible insights and apply AI to increasingly complex workflows.”

For Sabharwal, that was the ambition from the start: to reimagine information management in pharma and give teams a better way to develop medicines. “Nobody had ever managed to build an end-to-end lifecycle management platform for the pharma and biotech industry, but we made it happen,” he says.

With AWS providing a secure, scalable foundation, and Amazon Bedrock supporting flexible access to AI models, QbDVision is helping pharma organizations digitally transform their CMC operations — and building an AI platform with the potential to transform how new generations of medicines are developed and manufactured.