How NVIDIA’s AI Platforms are Reshaping the Future of BioPharma R&D

“We can cut preclinical development from five years to 18 months. Molecules which are AI-enabled in some way seem to be doing better in Phase 1 than standard molecules.” – Dr. Eva-Maria Hempe, Head of Healthcare & Life Sciences EMEA at NVIDIA

Dr. Eva-Maria Hempe leads healthcare and life sciences for NVIDIA across Europe, combining a background in physics with a PhD in healthcare services design to drive AI innovation across the pharmaceutical industry.

In the latest episode of the PharmaSource podcast, Eva-Maria explains how NVIDIA’s accelerated computing platforms are enabling breakthroughs in drug discovery, clinical trials, and personalised medicine through parallel processing power that’s revolutionising pharmaceutical R&D.

The Three Pillars of AI’s Pharmaceutical Revolution

As pharmaceutical companies embrace these technologies and approaches, Eva-Maria believes we’ll see dramatic improvements in drug discovery efficiency and effectiveness.

“AI development is exponential, but our brains are wired to think linear,” she says. “For the longest time, the exponential curve traces the zero line, and then suddenly it shoots up. We’ll have exponential knowledge and insight creation that will transform healthcare.”

The current AI transformation in pharmaceutical research is driven by three converging factors.

“The way I think about it is you need data, you need algorithms, and you need compute,” Eva-Maria explains. “With ChatGPT and then spreading out across lots of fields since then, all those three things came to maturity at the same time.”

This convergence has created a breakthrough moment for pharmaceutical AI applications, but Eva-Maria emphasises that continued progress depends on developing these three elements in sync. While companies have collected vast amounts of data, the quality and structure of that data remains crucial.

“Not all data is equally valuable,” she notes. “I sometimes call it the grandma’s attic problem. You think there must be something valuable in there and can spend an enormous amount of time and resources but you might find anything.”

Instead, forward-thinking pharmaceutical companies are taking a more strategic approach to data collection, conducting structured experiments specifically designed to fill gaps in their datasets.

Closing the Loop: The Lab-to-AI Connection Transforming Drug Discovery

One of the most exciting developments Eva-Maria discusses is the “lab in the loop” approach that connects AI systems directly to experimental laboratories, creating a continuous cycle of hypothesis generation and validation.

“The lab in the loop is this idea that you start with data, create a hypothesis, run an experiment, and feed that experimental data back into your database. This allows you to refine your hypothesis, creating a circle between wet lab and dry lab,” she explains.

This approach is already showing remarkable results. Eva-Maria cites a recent BBC article featuring a professor who was startled when an AI tool generated five hypotheses within 48 hours—including one he had been developing for years but hadn’t yet published.

The next frontier, according to Eva-Maria, is connecting these AI systems to automated laboratories. “If you think about this, it really allows you to create a lot more data than if you have to do everything by hand. You can do much better sampling, and that will be the next step.”

NVIDIA’s Toolbox: Accelerating Every Stage of Pharmaceutical Development

NVIDIA has developed a comprehensive suite of specialised tools for pharmaceutical companies, with the Clara platform encompassing multiple technologies targeting different areas of drug development:

• Parabricks: Software for sequencing of DNA and RNA data which accelerates genomic analysis from 30 hours to just 60 minutes
• MONAI Toolkit: Processes medical images for both human diagnostics and cellular analysis
• BioNeMo Framework: Handles biological language models for protein folding and molecule generation
• Holoscan: Enables real-time sensing and edge intelligence for laboratory applications

NVIDIA has recently released EVO 2, an AI foundation model for biomolecular sciences released in February 2025, in collaboration with Arc Institute with contributions from Stanford University and UC Berkeley

“It was trained on 9 trillion nucleotides from 100,000 species across the tree of life,” Eva-Maria says. “It can predict with 90% accuracy whether a previously unrecognized mutation would affect gene function or not.”

NVIDIA’s approach extends beyond just providing individual tools. They’ve created “blueprints”—reference architectures that combine multiple components into complete workflows. Eva-Maria likens these to Lego bricks that pharmaceutical researchers can assemble to create customised solutions.

“One example is a virtual screening blueprint. First, I need to know what my target looks like, so I use a protein folding model. Next, I need molecules to bind to that protein, so I use a molecule generator. Finally, I take both the protein and molecules and input them into a docking model to see which molecules fit best.”

Building Partnerships Across the Pharmaceutical Ecosystem

NVIDIA’s approach to supporting pharmaceutical innovation involves building relationships with companies across the entire spectrum of the industry. Eva-Maria explains that they engage with three main categories of partners:

“We work with everybody. We work with pharma, we work with startups, and we work also with a lot of other ecosystem partners,” she states.

For emerging biotech companies, NVIDIA offers Inception, a virtual accelerator program that provides benefits including access to educational courses, computing resources, and community support. “Inception is free to join and free to be in,” Eva-Maria notes. “We’re helping startups (including biotech, digital health, medev and others) to use our tech in the best possible way.”

The company currently supports over 4,000 healthcare & life sciences startups through this programme, giving special attention to those doing particularly groundbreaking work. This support helps early-stage companies overcome one of their biggest hurdles—accessing the massive computing power needed for modern AI applications.

For larger pharmaceutical companies, NVIDIA typically engages in two ways. First, they work through established partners like global system integrators (Accenture, Deloitte) and cloud providers that these companies already use. Second, they maintain direct relationships with a select group of “Lighthouse accounts”—companies with whom they develop deep strategic partnerships.

“It’s a two-way thing,” Eva-Maria explains about these partnerships. “We’re investing a lot of our time and resources, but we do this if we see a company on the other side who’s also ready to invest time and social capital, who really wants to drive a change, who’s willing to fight internal battles to change things.”

These deeper partnerships often involve centralizing a company’s computing resources, which Eva-Maria sees as crucial for success. “What I see with a lot of customers is it’s very distributed. Different units have their own compute. Nobody ever has enough compute, but nobody really has an overview of what’s going on. By centralizing this in a central AI platform team, you can create a step up for a company.”

Practical Advice for Implementing AI in Pharmaceutical Companies

For organizations looking to implement AI effectively, Eva-Maria offers several practical recommendations based on NVIDIA’s experience working with pharmaceutical companies of all sizes:

1. Centralize Your AI Strategy

Eva-Maria strongly advises creating a central AI platform team with proper prominence and authority. “It should sit somewhere where it has the prominence, the funding, and the ability to drive changes,” she emphasizes.

AI impacts every department—from R&D to marketing, HR, and finance—so positioning it as a central function rather than a technology subset helps ensure its transformative potential is realized.

2. Start with a Portfolio Approach

Rather than focusing exclusively on moonshot projects, Eva-Maria recommends a balanced portfolio of AI initiatives:

“It’s really about building a portfolio between things that create value fast and then also enable you to build both the skills as well as the stamina to tackle the stuff which is more exciting but probably further out.”

She draws a parallel with healthcare, where there are “interesting” use cases (like AI diagnostics) and “boring” use cases (like ensuring adequate supplies in operating theaters). The boring use cases often address immediate pain points and can build momentum for more ambitious projects.

3. Expect and Plan for Resistance to Change

Change management is critical when implementing AI. “It’s a very natural thing that people are worried about change,” Eva-Maria notes. “And this is not specific to AI—any sort of changes create resistance.”

The long timelines in pharmaceutical development make this particularly challenging. While AI may dramatically improve early-stage research, it could still take a decade for those benefits to result in approved drugs. Leaders need to build support for long-term transformation while delivering short-term wins.

4. Focus on Augmentation, Not Replacement

One effective change management strategy is to focus first on augmenting professionals rather than replacing their functions. Eva-Maria cites the example of ambient scribes in healthcare:

“The doctor isn’t on the keyboard trying to take notes while nodding, but actually having a real conversation with you, with AI listening in and summarizing it. That’s a fantastic use case. Healthcare professionals became doctors and nurses to help people, not to sit in front of a computer.”

By starting with applications that enhance rather than threaten professional identity, organizations can build positive momentum for AI adoption.

5. Invest in Education and Community

NVIDIA places significant emphasis on education and community-building to accelerate AI adoption. The company offers resources like the Deep Learning Institute and its Global Developer Conference (GTC) to help pharmaceutical professionals build their AI skills.

“There is a lot out there, and the community is very strong. That’s why we invest so much in the community—to help us better scale,” Eva-Maria explains.

For pharmaceutical professionals looking to explore these technologies first-hand, Eva-Maria highlighted NVIDIA’s upcoming GPU Technology Conference (GTC) taking place in San Jose from May 17-21, 2025. Originally a developer-focused event, GTC has evolved to include business-oriented sessions focusing on AI implementation and innovation strategies and will feature feature keynotes from NVIDIA CEO Jensen Huang alongside sessions with industry leaders including Nobel Prize winner Frances Arnold and representatives from major pharmaceutical companies like Pfizer and Roche.