GenAI Is a Strategic Challenge With Immense Opportunities — Merck Director for Data Sciences - Advanced Analytics

(US & Canada) Vijay Yadav, Director of Data Sciences-Advanced Analytics at Merck, speaks with Scott Shimp, VP, Solutions at Data Society, in a video interview about his team’s purpose at Merck, the challenges with generative AI (GenAI), educating stakeholders on GenAI, elements for delivering successful projects, the role of governance in GenAI projects, and the overall impact of GenAI on the Pharmaceutical industry.

Merck & Co., Inc. is one of the largest pharmaceutical companies in the world, generally ranking in the global top five by revenue.

At the outset, Yadav outlines that his team’s primary role encompasses the application of advanced technology to produce safe and quality medicines for patients worldwide. He adds that the challenge of applying AI and GenAI technologies becomes critical as the outcomes directly impact people’s health and lives.

Elaborating, Yadav maintains that the way Merck applies technologies such as GenAI internally within the company and externally by meeting regulatory compliance requirements directly impacts people. Keeping this in mind, the organization has started to delve deep into producing life-saving products.

Further, the spectrum is different when it comes to compliance, says Yadav, and ethical compliance is on top of that. He adds that it becomes critical to ensure the ethical use of AI and ML.

Generative AI is a strategic challenge

When asked if the emergence of GenAI shifted organizational focus, Yadav states that the company first envisions it as a strategic challenge. Considering data as a strategic asset, he adds that traditionally, organizations applied analytics on top of structured data sitting in databases, data warehouses, and data marts.

However, with the emergence of GenAI, organizations have to focus on unstructured data, which constitutes 75% of the total stored data, says Yadav. Now, the strategic challenge lies in bringing together the structured and unstructured data.

The next challenge, according to Yadav, is the need to have a solution developed for each GenAI use case in a way that is applicable with enterprise-grade scalability

Yadav shares that if organizations overcome these challenges, it will lead to a new level of GenAI insights. He wishes to develop a solution that has the entire knowledge base, including structured and unstructured data, so that it can meet any question in a specific domain.

Democratizing analytics has become easier now, says Yadav, and he adds that while generative AI is a strategic challenge, it also brings along immense opportunities.

Educating stakeholders on GenAI an organic journey

As GenAI continues to evolve, Yadav reflects on taking a structured approach to keeping stakeholders up to speed with the new technology. The journey has been organic, he says, and explains that there are different kinds of knowledge receptors in an organization.

First, there are the technical people, such as data scientists, who have a data background. The next kind are the business users who necessarily do not have an AI or data background, and the third group of people fall somewhere in the middle of the two extremes.

Taking the example of business users, Yadav notes that the company has focused on creating GenAI awareness and the discovery of use cases organically. He continues that the role of the data analytics team in educating business users involves training them on applying GenAI to individual situations.

In addition to awareness sessions, understanding the associated risks and ethics also forms a core part of training stakeholders from the beginning. Yadav reveals that after the initial session, the stakeholders are given 2-3 weeks to ponder and come back to start the use case discovery session. He shares that the process has been successful in building engagement, and his team has received numerous use cases from the business side.

The other side of the spectrum includes technical people, such as data scientists, who have foundational experience with AI. To further educate them on GenAI, they are put together to collaborate with people experienced in GenAI on projects.

It is a hands-on experience for this lot and is working well, as 90% of people are now GenAI-trained, says Yadav. He adds that as part of his team's efforts, they focus on a range of activities, including ongoing workshop training programs and boot camps.

The 4 elements of a successful GenAI project framework

To evaluate the success of GenAI projects, Yadav lists out four elements that constitute the evaluation framework:

1. Business Alignment 

2. Availability of Quality data 

3. Technical feasibility

4. Assessment of risk profile

Regarding business alignment, he states that all the projects must be well-aligned with the business. Next, Yadav affirms that without quality data, no fancy idea is worth taking up as a problem to solve.

Then, he stresses technical feasibility, wherein it is crucial to have the right resources to develop the solution. Highlighting the fourth element, Yadav maintains that, specifically in the pharmaceutical sector, where a decision impacts people’s health, it is a must to assess the risk profile.

Additionally, he suggests addressing all the elements and then narrowing the focus of the problem in a way that the organization can quickly reach for the low-hanging fruits to start showing value through the journey.

In continuation, Yadav sheds light on the different categories of projects that the company has looked into. First, he discusses how Merck is applying the search capability to get faster insights than manual processes. The searching capability is one category of solution with an open-ended framework, and the team has rebranded it as “intelligent search.”

The second capability is Q&A, which is applied the same way as solutions like ChatGPT, wherein a user can ask any question within a certain domain or subdomain.

The third category involves generating content, such as SOPs, based on certain criteria.

The role of governance in GenAI Projects and Merck’s approach to GenAI’s technical landscape

Emphasizing the aspect of governance with respect to GenAI, Yadav highlights the existence of the ethics and compliance committee at Merck. As a member, he affirms that all projects are filtered through a central team that evaluates the projects and then grants keys to develop GenAI solutions.

The evaluation focuses on ethical use, bias, and risks associated with the project elements. The team also examines the risk factors associated with the creation and use of synthetic data if real data is unavailable.

Apart from that, a part of governance also stresses the benchmark of the solution that is being developed and knowing how it is working. Yadav confirms that the company has developed some of its benchmarks internally, based on which models are assessed before rolling out.

Shifting gears, he comments on the technical landscape of GenAI, stating that Merck is at that level where it is fine-tuning and developing the models from the ground up. Therefore, the company leverages a mix of open-source and proprietary models, and it utilizes the Microsoft, OpenAI, and AWS platforms to do that.

Impact of GenAI on the pharmaceutical industry

Expanding on the GenAI adoption, Yadav analyzes its impact on the pharmaceutical industry, starting from drug discovery to research and development, and from manufacturing to reaching patients and understanding them.

Explaining further, Yadav shares a use case where a company utilizes failed clinical trial data to develop new molecules for medicines. Even with drug discovery, he points out how structured and unstructured data can be combined to do the research and development.

Furthermore, the application of GenAI reduces the clinical trial period. For instance, organizations can determine the effect of a certain medication based on prior real-world evidence about a person’s demography, age, and other conditions. Based on the insights, the effectiveness of the drug can be decided.

In the manufacturing space, Yadav shares how one of his teams has been working on a particular GenAI case that includes intelligent search, Q&A, and content creation. The idea behind this is to accumulate all the documentation related to pharma manufacturing saved as PDFs for a decade.

By gathering the knowledge in all the PDFs in a system, one can approach the system in case of any issues related to the manufacturing process and resolve those in minutes, says Yadav. When it comes to supply chain issues, GenAI can be leveraged to predict any impending risk and take proactive action.

Thereafter, Yadav remarks on how pharmaceutical industries can use the patient feedback data collected from different channels and use that information to improve the next product. In conclusion, he asserts that GenAI has impacted the entire value chain of the life science industry.

CDO Magazine appreciates Vijay Yadav for sharing his insights with our global community.