Ivan Kairatov is a prominent biopharma expert with a deep understanding of how technological innovation reshapes research and development. Currently, he is focusing on the intersection of artificial intelligence and oncology, advocating for a transformative approach to medical education. Through his work, he emphasizes the necessity of “bilingual” researchers—professionals who can navigate the complex languages of both molecular biology and computational science to unlock personalized therapies.
This interview explores the evolving landscape of cancer research, focusing on the integration of massive genomic datasets with large language models. We discuss the logistical framework of dual-track training programs, the ethical oversight required when handling synthetic data, and the growing alignment between academic training and pharmaceutical industry standards.
Traditional research often separates biology from computation. How does a dual-track training program bridge the gap between these disciplines, and what specific skills should clinical fellows prioritize to effectively use large language models for molecular characterization?
The traditional siloed approach is no longer sufficient because cancer research now involves massive datasets spanning genomics, imaging, and clinical outcomes that only AI can fully process. Our dual-track program bridges this gap by training clinical oncology fellows to become “bilingual,” meaning they can speak the language of both biology and computation. The learning process begins with foundational literacy in AI large language models, followed by hands-on training in how these models can molecularly characterize a tumor the moment a patient is diagnosed. Fellows must prioritize the ability to synthesize vast amounts of existing knowledge to identify customized therapy options quickly. Finally, the training culminates in the practical application of these tools to deliver on the promise of personalized cancer therapies in a clinical setting.
Effective training often involves pairing experts from different fields. What are the logistical challenges of matching a computational mentor with a cancer biologist, and how does this collaborative structure directly impact the speed of developing personalized therapies?
The primary logistical challenge lies in synchronizing two very different research cultures and ensuring that the computational mentor and the cancer biologist share a unified vision for a trainee’s progress. We address this by assigning each trainee to a specific mentor pair, creating a collaborative triangle that forces constant communication between the lab and the data terminal. This structure directly impacts the speed of therapy development because it eliminates the lag time typically lost in translation between data scientists and clinicians. By having both perspectives present from day one, we can more rapidly move from a genetic discovery to a clinical trial design. This collaborative model ensures that the computational tools are always answering the most urgent and relevant questions for the patient.
With the rise of synthetically generated data and potential privacy risks, what oversight mechanisms are necessary for researchers using AI? How can institutions teach the next generation to detect inaccurate results while maintaining ethical standards?
As a “super user” of AI, I have seen firsthand that these powerful tools require careful supervision to avoid ethical breaches, such as the creation of fake papers with synthetically generated data. We implement strict oversight mechanisms that focus on human-in-the-loop validation, where every AI-generated result must be cross-referenced with established biological benchmarks. To protect patient data, we utilize secure institutional infrastructures and specialized “AI clinics” that offer workshops on the safe analysis of medical records. We teach our students that detecting inaccuracies is just as important as generating results, instilling a culture of skepticism and rigorous verification. Maintaining ethical standards is not just about following rules; it is about training researchers to understand the weight of the data they handle.
Pharmaceutical companies are increasingly using AI to design clinical trials and monitor drug safety. How can academic training programs keep pace with these industry shifts, and what practical steps should researchers take to ensure their AI tools meet regulatory requirements?
Academic programs must evolve rapidly because pharmaceutical companies are already deploying AI agents to handle complex tasks like monitoring drug safety and ensuring regulatory compliance. To keep pace, we are integrating industry-standard tools into our curriculum so that our researchers are prepared to work in these high-stakes environments upon graduation. Researchers should take practical steps by engaging with institutional initiatives, such as the AI to Advance Medicine program, which provides the necessary services to ensure AI adoption is safe and effective. It is vital to bridge the gap between academic discovery and industrial application to ensure that no generation of scientists is left behind. This integration ensures that the tools developed in the lab are robust enough to withstand the scrutiny of federal regulatory bodies.
Building institutional infrastructure like specialized AI clinics requires significant resources. What are the most critical components for setting up a secure environment to analyze medical records, and how can these clinics effectively train faculty members who are unfamiliar with computational tools?
The most critical components of a secure infrastructure are centralized data services and a robust framework for ethical AI adoption that faculty, staff, and students can all access. Our AI clinics serve as the frontline for this, offering both in-person and remote training sessions via platforms like Zoom to meet faculty where they are. The implementation strategy involves starting with targeted workshops—for example, focusing specifically on how to securely use AI to extract information from medical records. By providing this hands-on, accessible training, we demystify computational tools for veteran researchers who may feel intimidated by the technology. This phased approach allows the institution to build a culture of AI fluency while maintaining the highest standards of data security and patient privacy.
What is your forecast for the AI revolution in cancer research?
I believe we are entering an era where AI will become as fundamental to oncology as the microscope once was, leading to a future where every patient receives a truly customized treatment plan within days of diagnosis. We will see a shift where federal, private, and institutional investments focus heavily on creating “bilingual” scientists, as this human-AI partnership is the only way to navigate the complexity of modern genomic data. In the coming years, AI will not just analyze data but will actively suggest novel drug combinations and predict patient responses with unprecedented accuracy. The ultimate success of this revolution will depend on our ability to train a workforce that is ethically grounded and technically proficient. My forecast is that this integration will significantly close the gap between clinical research and successful patient outcomes.
