Interview with Shinya Yuki, CEO and co-founder of Elix.

Shinya-san, can you tell us about your professional background and your current role at Elix, Inc.

I earned my Ph.D. in physics, where I focused on astrophysics simulations using supercomputers. While I still love physics, I wanted to broaden my scope and work more directly in computer science. After graduation, I became a software engineer and worked in both Tokyo and Singapore. During that time, the deep learning revolution was beginning, and I was fascinated by its potential. I started learning on my own and eventually decided to found my own AI company.

In the beginning, we took on a wide variety of projects where AI could be applied, since I was still exploring the most meaningful use cases. One of those projects was in drug discovery, and I quickly realized how impactful AI could be in this field. Drug discovery not only has enormous societal value, but it also connects areas of science I am deeply interested in, from physics to chemistry to biology. That curiosity and sense of purpose made it clear this was the right direction for us.

Today, as CEO of Elix, I oversee the entire company, which is now fully dedicated to advancing drug discovery through AI.

Elix recently commercialized an AI drug discovery platform using federated learning with data from 16 pharma companies. What are the biggest opportunities and challenges you see in scaling this model while ensuring trust and data security?

Federated learning represents a tremendous opportunity in drug discovery. AI is a data-driven approach, and the more data you have, the better the model you can build. If you want to create the most powerful AI models, the boldest idea would be to aggregate all data from all pharmaceutical companies. In reality, that is impossible because of data confidentiality, and it would even sound unrealistic to suggest. What federated learning enables is the ability to train across data from multiple companies while still keeping that data confidential.

Security is always a major concern for pharmaceutical companies, but with federated learning, the data never leaves the company’s environment. At Elix, we developed our own federated learning framework called kMoL™, built from scratch by our engineering team, which makes it possible to train high-quality models in this collaborative setting.

However, technical solutions are only part of the challenge. Bringing together 16 pharmaceutical companies is an enormous task in itself. Each company has its own priorities, contracts must be negotiated, and not everyone is deeply familiar with the details of machine learning or federated learning. There is also the question of incentives. In any community, there can be free-rider issues: if some companies share less data but still benefit from a model trained with contributions from others, it can feel unfair. Creating a fair and sustainable incentive structure has been essential.

Overcoming both the technical and business challenges has not been easy, but that is also where the strength of this model lies. Because it is difficult to build and to manage, it is not something that can be easily copied. The very challenges we faced have become part of our moat, and successfully addressing them has created a strong foundation for scaling this initiative.

Elix Discovery™ being adopted by KYORIN and Eisai marks a major milestone. Could you share the strategic vision behind achieving adoption at this scale, and what you believe made KYORIN and Eisai choose Elix’s platform?

Our work on federated learning with 16 pharmaceutical companies laid an important foundation. Through that collaboration, we had already built strong relationships and trust with many companies even before commercializing Elix Discovery™. The AI models developed through federated learning are now integrated into our platform, so for many partners, adopting it was a very natural next step.

Another important factor is scale. In Japan, we believe Elix Discovery™ now holds the largest share among AI drug discovery platforms. This creates a strong network effect. For example, we host a user group where pharmaceutical companies can exchange knowledge, share best practices, and learn from each other. The more users we have, the more value the entire community gains.

What also sets us apart is our commitment to dedicated support. Other platform providers mainly deliver the software itself without extensive follow-up. In contrast, we provide comprehensive support, including educational calls that cover everything from the basics of machine learning to practical guidance on how to use the platform effectively and generate promising molecules. By working closely with medicinal chemists, who often do not have a machine learning background, we make sure the technology is not just available, but truly usable in their day-to-day research.

Your platform is designed under the principle that ‘medicinal chemists can truly use it.’ As CEO, how do you ensure this user-centric philosophy continues to guide Elix’s product evolution?

From the very beginning, we have prioritized medicinal chemists as the primary users of our platform. Of course, computational chemists can also benefit from Elix Discovery™, but the majority of researchers in pharmaceutical companies are medicinal chemists, so it makes the most sense to design with them in mind.

When we first engaged with pharmaceutical companies, we heard a consistent message. Computational chemists often have small teams but must collaborate with many medicinal chemists. They told us that if a platform is designed primarily for medicinal chemists, it would ultimately make their own work easier as well. By contrast, many products in the market are built with computational chemists in mind, and while they can technically be used by medicinal chemists, they are not truly designed for them.

Our approach has been different from the start. Most of our users are medicinal chemists, and their feedback continuously informs how we evolve the platform. We also have medicinal chemists within Elix, whose input helps us refine the interface and ensure it is practical for day-to-day research. By consistently gathering feedback from both external and internal chemists, we keep our platform aligned with the needs of medicinal chemists and uphold our philosophy that they can truly use it.

Elix Discovery™ combines ligand-based, structure-based design, AI consulting, and implementation support, all with an intuitive GUI. How do you prioritize these elements to stay competitive and aligned with client needs?

There are many directions in which we could expand Elix Discovery™, but our priorities are guided by client feedback and insights from collaborative research. When new features prove essential in advancing a joint project, and we see that they would also benefit other users facing similar challenges, we integrate them into the platform.

In drug discovery, both ligand-based and structure-based approaches are important, which is why our platform supports both. For example, in some projects, clients already have biological activity data but lack structural information on the target, which makes a ligand-based approach more effective. In other cases, there is no biological activity data, but structural information is available, in which case structure-based approaches such as docking simulations become critical.

We continuously review and decide on development priorities through regular discussions, often on a monthly basis, to ensure that the most useful and needed features are addressed first. This process not only helps us respond directly to client needs but also ensures that we remain competitive by focusing on the features that deliver the greatest value to drug discovery projects.

In commenting on the Shionogi collaboration, you pointed out that previously, retrosynthetic AI models lacked practicality due to algorithmic complexity and limited data. How has Elix addressed those limitations, and what role do strategic partnerships play in that?

Retrosynthesis models are in high demand, but there have traditionally been two major challenges to making them practical: algorithms and data. Compared to property prediction, retrosynthesis is more complex and requires deeper expertise in machine learning to develop reliable models. This is an area where Elix has strong capabilities.

The other challenge is data. Publicly available datasets mostly come from patents, and while useful, they often lack the quality and scientific rigor needed for robust model training. On the other hand, pharmaceutical companies have access to high-quality reaction data but may not have the resources or expertise to build and maintain complex AI models themselves.

This complementary set of strengths is exactly why collaboration is so powerful. At Elix, we can focus on developing advanced AI models, while our pharmaceutical partners provide the high-quality reaction data needed to train them effectively. By combining these capabilities, we can overcome the limitations that have held back retrosynthesis models in the past and make them far more practical for real-world drug discovery.

Elix has a growing track record of collaborations, from Astellas Pharma (2020) and Shionogi (2022) to strategic work with and others. As CEO, what is your overarching strategy for selecting and cultivating these partnerships?

Our strategy for partnerships is guided by the principle of complementary strengths. We collaborate when we see that Elix can make a meaningful contribution and, at the same time, benefit from what our partners bring. A typical case is that we provide advanced AI models, while our partners provide the high-quality experimental data needed to train them.

As an AI company, our strength lies in developing sophisticated models, but we do not generate experimental data ourselves. Pharmaceutical and biotech companies, on the other hand, produce data through their research and experiments but may not have the resources or expertise to build complex AI models. Working together allows both sides to overcome their limitations and create more effective solutions.

In some cases, we also support computational chemists within pharmaceutical companies. Even though they may already be familiar with machine learning, they are not always able to dedicate the time, resources, or specialized expertise required to develop complex models. By partnering with them, we help extend their capabilities and accelerate their research.

With Elix’s origins in deep learning and life sciences, and with Blockbuster Tokyo and the Life Intelligence Consortium, how do you view Elix’s positioning today in the evolving AI drug discovery landscape?

When I first started Elix, we were very much a pure AI company and not yet familiar with drug discovery. The acceleration program, Blockbuster Tokyo, was valuable at that stage because it allowed us to receive feedback from domain experts and better understand the realities of the field.

The Life Intelligence Consortium was also a turning point. It brought together academia and pharmaceutical companies, and through that community we were able to learn what truly matters in drug discovery while also building trust with key stakeholders. That trust was essential in enabling us to start working on projects with pharmaceutical companies, and it ultimately allowed us to evolve from a general AI company into one focused on drug discovery.

The experience and relationships we gained through these initiatives continue to be a strength today. They helped create the foundation that later enabled us to unite 16 pharmaceutical companies and commercialize a federated learning project, something that would have been impossible without the trust built in those early days.

You’ve previously shared insights through lectures (e.g., with Tokyo Institute of Technology, Chem-Bio Informatics Society) and media outlets like . How does this engagement with academic and industry audiences influence Elix’s strategic direction and your approach as CEO?

Engaging with both academic and industry audiences is an important part of our strategy at Elix. For example, we regularly participate in conferences such as the Chem-Bio Informatics Society annual meeting, where both academic researchers and pharmaceutical companies come together. We often give lectures, oral presentations, or sponsor sessions at these events. This is essential because many of our collaborations are with pharmaceutical companies and academia, and these forums allow us to exchange ideas, build trust, and strengthen our reputation in the community.

On the academic side, I also lecture at universities, such as at Tokyo Institute of Technology, now known as Science Tokyo. I often speak to international students, sharing not only technical insights but also my experience of building an AI drug discovery company from the ground up. Education is meaningful to me, and if my story can inspire students as they think about their own careers, I am glad to contribute.

These activities reinforce Elix’s positioning by deepening our connections with both academia and industry, while also reminding me, as CEO, of the broader impact we can have beyond our immediate business goals.

Elix began as a consulting and platform company but has indicated moves toward platform licensing and collaborative research projects. How are you balancing these two business models, and what do you see as the future of each in your corporate roadmap?

When I first started the company, our work began as consulting projects with pharmaceutical companies. Over time, we noticed that many of these projects shared similar needs. That insight led us to develop an AI platform, because creating a reusable, scalable solution was far more efficient than building something similar again and again. A platform could deliver greater value to a wider range of users.

Today, our business is built on two models: platform licensing and collaborative research. The platform licensing business is subscription-based, which gives us stable and predictable revenue. This stability is essential for the long-term sustainability of the company, even if the growth rate of this business is somewhat limited.

Collaborative research is very different. Drug discovery projects are inherently risky, and many do not succeed. But when they do succeed, they create the potential for very significant returns. This makes collaborative research a high-risk, high-reward model.

Our strategy is to balance these two approaches: leveraging platform licensing to ensure stability, while also pursuing collaborative research to capture opportunities for major breakthroughs and growth. This combination allows us to remain both resilient and ambitious in our roadmap.

Looking ahead, how do you envision Elix shaping the global biotech and pharmaceutical landscape? What do you see as the key trends, and how is Elix positioning itself to lead or adapt in that future?

When I founded Elix in 2016, there was much more skepticism around AI. The evidence of its effectiveness in drug discovery was limited, and many people were unsure whether it could truly make an impact. Today, the landscape is very different. Advances such as large language models have made AI part of everyday life, and people now see its usefulness and potential more clearly. Even the Nobel Prize has recognized AI research, which shows how far the field has come in gaining global credibility.

In drug discovery, this shift means that nearly all pharmaceutical and biotech companies now recognize the importance of AI. However, awareness alone does not solve the challenge. Many companies do not have the AI engineers or researchers needed to fully take advantage of these technologies. That is where Elix plays a critical role, helping bridge the gap between AI innovation and practical application in drug discovery.

Our federated learning project is a good example of this. Collaborating with 16 pharmaceutical companies around a shared AI framework was a major achievement, and it has strengthened our position as a trusted partner in the industry. Building on this foundation, we believe Elix is well-positioned to continue leading in the global biotech and pharmaceutical landscape, both by advancing AI technologies and by ensuring they are accessible and practical for the organizations that need them most.

As more pharmaceutical and biotech companies explore AI in drug discovery, what message would you share with leaders in the same space about embracing innovation while navigating challenges?

AI is a transformative technology, and we are now seeing more and more case studies that demonstrate its effectiveness in drug discovery. This makes it clearer than ever that investing in AI is the right direction for the industry. At the same time, pharmaceutical and biotech companies are not always equipped to make full use of these technologies on their own. In those situations, seeking support can be very valuable. At Elix, we are committed to helping our partners apply AI effectively and meaningfully in their research, and we are always open to collaboration.

Thank you, Shinya-san!