The traditional 7–10-year timeline for drug discovery is rapidly becoming obsolete. Thanks to advances in AI and compute infrastructure, pioneers like Nebius are spearheading a shift that compresses this cycle to as little as 18 months. In this exclusive BioPharma APAC interview, Dr. Ilya Burkov shares how Nebius’ full-stack, AI-native platform empowers scientists, clinicians, and biotech innovators to simulate, test, and validate drug candidates faster and more affordably than ever before. From transforming early-stage development to supporting rare disease research in Asia, AI is no longer an experimental edge—it’s a foundational force for the future of precision medicine.

AI is helping compress the traditional 7–10-year drug discovery cycle to as little as 18 months. How is Nebius enabling this shift, and what do you see as the most transformative stage of the drug development journey that AI currently impacts?
 

Innovation in healthcare has traditionally been very slow, expensive, and bogged down by clunky infrastructure. Nebius is actively transforming this by providing scalable, high-performance AI infrastructure that significantly accelerates drug discovery and development. 

Our vertically integrated platform enables biotech innovators, researchers, and scientists to achieve in hours what once took years. For example, Converge Bio used our AI platform to analyze 36 million single cells in days, a process that would typically take months, which was crucial for understanding disease mechanisms at a cellular level. Similarly, Simulacra AI leverages Nebius' distributed compute infrastructure to build wave-function foundation models for molecular systems, generating high-accuracy datasets for drug and material discovery pipelines. This allows for virtual testing and 'failing faster' in a simulated environment, quickly ruling out unpromising drug candidates and focusing resources on those with the highest chance of success.
 
One of the most transformative stages AI currently impacts is early-stage drug development, particularly target identification and compound screening. It automates stages processes that traditionally took months or years by generating novel drugs with a high likelihood of meeting specific criteria, like binding to a particular receptor while having low toxicity. 

This significantly narrows down the list of promising candidates, reducing early-stage development from years tomonths. Al-driven models are expected to be involved in the discovery of up to 30% of new drugs by the end of 2025. Our platform, including Nebius AI Studio, offers access to powerful GPUs and developer-friendly tools to train, run, and scale advanced models without the usual cost and complexity.

Could you elaborate on how AI is evolving clinical decision-making today—and what clinicians should be doing to stay ahead?

AI is not about replacing human judgment but augmenting it to enhance efficiency, accuracy, and ultimately, patient care. Clinicians today face a deluge of data, from complex imaging scans to vast electronic health records. AI acts as a 'diagnostic co-pilot', sifting through immense datasets to provide unbiased second opinions, reduce diagnostic errors, and standardize care quality. 

For example, a recent development from Northwestern University  saw AI tools trained on millions of images matching the accuracy of doctors in outlining lung tumors on CT scans and flag high-risk areas that might be missed. This frees up radiologists to focus on the most complex aspects of a case, consult with colleagues, and, most importantly, communicate with patients.

Beyond diagnostics, AI streamlines workflows and reduces administrative burdens. Physicians often spend two hours on paperwork for every hour of patient interaction. AI scribes, by transcribing consultations in real-time, save clinicians hours of administrative work, allowing them to be more present and engaged with patients. This reclaims time for the invaluable 'human touch' that AI cannot replicate.

To stay ahead, clinicians must embrace AI literacy and integrate these tools into their practice as early as possible. This means understanding AI's capabilities and limitations, how to interpret its outputs, and how to effectively use it to enhance patient outcomes. Investing in training and participating in interdisciplinary discussions on digital health policy and governance are crucial steps. The future clinic will be led by clinicians empowered by AI, delivering more precise, personal, and profoundly human care.
 

With each rare disease affecting only a small number of patients, data scarcity is a known barrier in research. How can AI help unlock meaningful insights from such limited datasets, especially in the context of precision medicine in Asia?

AI provides a powerful solution for gaining valuable insights from limited rare disease data, crucial for personalized treatments in Asia.

Advanced AI models can now generate realistic, synthetic datasets, simulating thousands of 'virtual' patients. This is vital for rare diseases, allowing researchers to train AI and test ideas without compromising privacy. In Asia, with its diverse genetic populations often underrepresented in global data, this enables building locally relevant precision therapies. Combined with AI models analyzing individual cells and multiple biological data types, we can uncover previously unattainable insights.

AI is unlocking value from limited data through smarter algorithms, synthetic augmentation, data harmonization, and privacy-preserving collaboration. For rare diseases, especially in the Asian context, AI is not only helping to fill in knowledge gaps but also accelerating the development of targeted therapies and diagnostic tools tailored to diverse populations.

For clinicians and researchers in Asia, the call to action is clear: embrace AI as a partner, not a panacea, and work toward building inclusive, collaborative, and high-quality datasets that ensure no patient is left behind simply because their condition is rare.

What’s your perspective on the hidden bottlenecks in cloud compute for life sciences, and why is simply having GPU access no longer enough?

The complex, specialized nature of AI models and massive biomedical data require a purpose-built, full-stack AI infrastructure. Generic cloud offerings are inefficient, leading to operational overhead and higher costs for researchers trying to adapt general-purpose tools for highly specific scientific tasks. 

For instance, training an AI model to analyze vast genomic sequences or simulate complex molecular interactions demands not just raw processing power but also specialized software, optimized data pipelines, and robust security features tailored for sensitive health data. Innovators need environments optimized for scientific computing and secure handling of sensitive data, which often isn't a core offering of general cloud providers.

For example, SieveStack uses a special cloud platform built for AI to create very detailed simulations of molecules and train powerful AI models. This helps them use over 90% of their computing power, making drug discovery much faster by reducing delays and speeding up training by as much as four times.  
 

This is where platforms like Nebius become critical, as they are specifically designed to provide the comprehensive, AI-native environment required for life sciences breakthroughs, allowing innovators to focus on disease research and not server management.

 
As we’re still in an “experimental” stage with AI in healthcare, from a regulatory and clinical validation standpoint, what are the most urgent guardrails that need to be in place to ensure responsible scale-up?

Firstly, a unified AI governance framework for health is needed to address regulatory uncertainty and ensure safety, essentially a clear rulebook for how AI can be used in medical settings that fosters trust and clarity for developers and users alike. Strong data governance and privacy laws are also paramount to prevent misuse of sensitive patient data and ensure model trustworthiness. This entails strict rules on how patient health information is collected, stored, and used, given the immense volume of data AI consumes.

Setting up standards for AI model validation and clinical safety are essential to evaluate efficacy and bias, ensuring AI tools are proven safe and effective before widespread use and preventing potentially harmful decisions. Clear procurement and reimbursement policies are necessary for AI tools to scale in healthcare systems, defining how these tools are purchased and paid for (e.g., insurance coverage for an AI-driven diagnostic test), which is vital for broader adoption. 

Finally, capacity-building in AI literacy for clinicians and regulators is vital to ensure proper tool usage and prevent errors, meaning training healthcare professionals to understand, interpret, and work effectively with AI as a complementary technology

Looking ahead to the next 3–5 years, what role do you see Singapore playing as an AI-powered biotech hub in Asia Pacific—and what will it take to realise that vision?

Singapore is exceptionally positioned to become a leading AI-powered biotech hub in Asia Pacific over the next 3-5 years, building on its strong foundations like the National Precision Medicine Programme and RIE2025. Its government-led AI strategy, research focus, and business-friendly environment provide an ideal base for serving Southeast Asia.

Singapore will serve as a critical launchpad for specialized AI solutions in biotech and healthtech, moving beyond general AI to areas like diagnostic intelligence and personalized therapies. Its commitment to regulatory sandboxes, such as IMDA's AI Verify, will ensure responsible innovation.

Realizing this vision requires several key factors. Equitable access to scalable AI infrastructure, including affordable, high-performance computing like GPUs, is paramount for all startups and researchers. Continued public-private partnerships are essential to provide the necessary infrastructure, technology, and expertise, accelerating data interoperability and responsible AI adoption. 

Furthermore, nurturing specialized AI talent through continuous upskilling and reskilling initiatives is also crucial to ensure a skilled workforce. Fostering a data-driven ecosystem that encourages ethical sharing and integration of diverse genomic and clinical datasets will unlock advanced insights for precision medicine. 

Finally, decentralizing innovation is important. This means enabling emerging ecosystems across Southeast Asia with scalable cloud tools and open-source models to help amplify the region's overall impact and create a vibrant, interconnected network. By focusing on these areas, Singapore can solidify its leadership, drive economic growth, and improve health outcomes across Asia Pacific.