Late last year, Quotient Sciences, a global drug development, research and manufacturing accelerator, and Intrepid Labs, an AI leader in pharmaceutical formulation science, announced a multi-year strategic partnership to advance the use of artificial intelligence in early drug development.

The partnership leverages Quotient’s established capabilities in integrated drug product development, manufacturing, and clinical testing, alongside Intrepid’s machine learning model, ANDROMEDA, an AI platform designed to develop and optimize clinical performance of drug products.

Contract Pharma caught up with Andy Lewis, Chief Scientific Officer at Quotient Sciences, and Christine Allen, CEO and Co-Founder of Intrepid Labs, at DCAT Week, where they elaborated on the strategic partnership and shed light on their joint mission to accelerate drug product development through the integration of advanced AI and machine learning technologies.

Origins of the Partnership

This latest partnership builds on an existing collaboration where Intrepid’s AI model was incorporated into Quotient’s Translational Pharmaceutics platform to help accelerate the identification of optimal formulation compositions and reduce the time to transition new drug products into clinical development with even greater chances of success. 

“At Intrepid, our goal has always been to integrate automation and AI to accelerate drug development. We have a very strong machine learning team—Toronto is a major center of excellence for AI—but AI alone isn’t enough. You need deep domain experts who can shape and refine the algorithms within real-world constraints,” explains Christine Allen.

According to Allen, partnering with Quotient enabled Intrepid to combine its AI-first technology with a global CDMO that brings more than 30 years of drug development expertise.

“Their Translational Pharmaceutics platform is already one of the best in the industry, used across more than 500 drug development programs for over 100 clients,” Allen says. “Our aim was to help make something excellent even better.”

“Quotient Sciences has long envisioned how AI and machine learning could accelerate and enhance Translational Pharmaceutics,” says Andy Lewis. “I’ve known Christine, who founded Intrepid Labs, for years. When she launched the company, she reached out—and it immediately became clear that there was strong synergy between our capabilities.”

He explains, “At Quotient Sciences, we integrate drug product development, manufacturing, and clinical testing. This allows us to reduce API demands in early development, speed up entry into the clinic, and improve decision making through just-in-time manufacturing.”

Intrepid Labs’ proprietary machine learning platform, Andromeda, offers similar advantages, says Lewis.

“It enables rapid, efficient identification of formulations that meet predefined targets, again using very little API. At the same time, it builds an in silico model of the drug product, which supports more informed decision making throughout development,” he explains.

After working together for two years, Lewis claims the results are compelling: “We can save up to 50% of formulation development time and achieve a comparable reduction in API usage. The in silico predictive model is especially powerful.”

What Sets this Partnership Apart

Many CDMOs are investing in AI right now, but Lewis believes Quotient’s work with Intrepid is unique. He claims, their solution is the only AI-driven engine focused specifically on optimizing the clinical performance of formulations.

“Our solution uses an active learning approach with Bayesian optimization, which allows it to learn from sparse datasets. That means we don’t need thousands of data points to build effective models. This is crucial early in development, when data is limited,” says Lewis.

“Other AI approaches, such as neural networks, typically require very large datasets and therefore don’t perform as well in early-stage formulation development,” he explains. “Our approach works from the start, adapts as data grows, and directly optimizes for clinical performance.”

About the ANDROMEDA Platform

Both Lewis and Allen underscore that ANDROMEDA is not a large language model, but an isolated, low-compute, project-specific algorithm designed to protect client data and operate effectively with sparse datasets.  

“Large language models, like ChatGPT, are trained on enormous datasets that include a lot of third-party information. Andromeda is entirely different,” explains Allen. “It uses active learning and Bayesian optimization, not large language model architecture.”

What this means is ANDROMEDA is very low compute—you could run it on a laptop—which makes it cost-effective and energy-efficient. When Quotient and Intrepid start a new client project, they create a fresh instance of the algorithm that has not been trained on any third-party data. All data and IP generated belong solely to the client. After the project ends, the instance is either destroyed or securely stored.

According to Allen, “This gives clients confidence that their data will not be used to benefit anyone else. The isolation and control are built into the model from the start.”

Why AI is Taking Off Right Now

Allen calls the release of ChatGPT in 2022 “a global turning point.” She says, “Suddenly, everyone could see how AI could support and enhance human decision-making.”

“Given that more than $100 billion is spent globally each year on drug R&D—yet 90% of molecules still fail in clinical development—the need for better decision making is clear,” she remarks.

Furthermore, Allen claims, “AI can examine a much broader design space and optimize for a larger number of objectives simultaneously. It helps minimize API use, reduce experiments, and identify formulations more efficiently, ultimately increasing the likelihood of success.”

“AI has enormous potential to improve the quality and efficiency of drug development,” adds Lewis. “For example, a recent review showed that 65–75% of FDA complete response letters highlight deficiencies in CMC or manufacturing. There is a tremendous opportunity for AI to address those challenges.”

Looking Ahead: AI’s Potential Impact on Formulation Science

“In our work with Quotient, we’ve shown that AI can significantly accelerate the identification of optimal formulations,” says Allen. “It can identify non-obvious formulation options that human experts, even highly experienced ones, might not consider. And it generates comprehensive datasets that map the relationships between composition, properties, and performance.

“This is incredibly valuable during scale-up and manufacturing. If you need to adjust an excipient level, the algorithm can predict how that change will affect performance. So, AI doesn’t just help at the early stage, it supports decision making throughout development.

“As we continue demonstrating time and cost savings and improved outcomes, I expect adoption in formulation science to grow rapidly.”

Lewis expects AI-enabled control of formulation and manufacturing processes to become standard over the next 5–10 years.

“This will improve right first-time success, enhance product quality, and significantly shorten development timelines, ultimately getting innovative medicines to patients faster,” he says.

“It’s a very exciting time,” concludes Allen. “I encourage any AI-first technology company to partner closely with organizations that have deep domain expertise and a strong understanding of real-world constraints. There’s a fear that AI will replace experts, but I think the real mistake would be not involving them. The best outcomes come from combining advanced machine learning with decades of human expertise. Our work with Quotient is a great example of that synergy.”