As a PhD student, Hetu Kamisetty was drawn to the potential of artificial intelligence (AI) to advance biological research, particularly in the design of new drug treatments. However, at the time, “the technology wasn’t there,” he said.
Now, nearly two decades later, Kamisetty leads the AI and machine learning efforts as the Chief Technology Officer at Xaira Therapeutics, a biotech startup launched in 2024 with a staggering $1 billion in venture capital — one of the largest funding rounds in recent history.
Co-founded by Kamisetty’s former postdoctoral advisor, the Nobel Laureate David Baker, the team at the startup seeks to build on Baker’s RFdiffusion and RFantibody models, which researchers can use to design functional proteins and antibodies from scratch (1,2).

As the co-founder and Chief Technology Officer of Xaira Therapeutics, Hetu Kamisetty leads AI and machine learning efforts at the biotech startup.
Credit: Xaira Therapeutics
Xaira hopes to leverage these models to develop new molecules, connect them to specific biological targets, and identify patients who might benefit from a particular treatment. Their goal is to create new medicines for diseases that have been historically difficult to treat.
How did you become interested in using AI for drug discovery?
I’ve always been interested in building models to explain parts of the world, and during my PhD, I found it fascinating how many unsolved problems there were in biology. This led me to do postdoctoral research in David Baker’s lab after I graduated in 2010. That experience really sparked my interest in the space of protein design and structure. But, at that time, the field was nascent, and the technology that would have allowed us to do what Xaira is setting out to do today just wasn’t there. Now, the science around designing molecules has matured to the point where it has become more of a technology because of AI. That’s why we think this is the right time to be working on accelerating progress in drug discovery. A lot of problems that were not solvable before AI are now solvable, and that’s what we’re excited about at Xaira.
What motivated you to start Xaira Therapeutics?
In the last four to five years, it became clear to us that the pace of scientific breakthroughs translating into technological platforms was accelerating. For instance, only a few years ago when we started actively thinking about Xaira, we didn't yet have the state-of-the-art, generative AI models like chatGPT and Diffusion that we now take for granted. But it was very clear where the space was going, and that future improvements were needed to unlock that potential. Now it's become more obvious to everyone that generative AI is here to stay. The problems were there; the toolkit was emerging but nascent; and the people were there. Then it became the right time.
What are Xaira’s goals as a company?
There are three core challenges in drug discovery. We first have to identify the right causative mechanism for disease, then build a molecular entity that corrects that mechanism, and finally target it to the right set of patients to have the outcome we want. At Xaira, we are interested in all three aspects. We would like to make molecules for targets that have not been historically targetable, and we would like to better customize treatments by finding the right groups of people that will respond best to the therapies we’re designing.
What has surprised you most about the AI and drug discovery field?
The learning that continues to remain, and how much opportunity there is to change the status quo. There are a lot of areas where we could accelerate scientific progress in drug discovery, and we’re still scratching the surface. We’re constantly learning about new approaches, and hopefully that remains the case in the years to come. For example, a few years ago, I knew about the need to design new molecules. I knew about the problem of finding the right target. But I hadn’t really thought deeply about the patient stratification problem. It was only as the group that eventually became Xaira began to emerge that I learned the full magnitude of this problem and the opportunity there. Now it’s one of the central pillars of our company.
Are there any ethical considerations you think should be addressed in the use of generative AI for drug discovery?
Unlike other uses of generative AI, the models we train and build are based on scientific data. The drug discovery process itself has an exceptionally robust regulatory mechanism, so that regardless of how we make a drug, regardless of the ingredients that go into accelerating its development, it is held to the same scientific standards to demonstrate safety and efficacy. I can't think of ethical issues beyond these kinds of things because they would always be front and center for us: the safety and efficacy of the medicines we make.
What’s next for Xaira?
We are making medicine, and we are 100 percent focused on making medicine. We believe making medicine is an extremely valuable endeavor for society, and an extremely valuable scientific achievement. And obviously, if we do it right, it will also be good business. But we've just gotten started, so there's a lot to do.
This interview has been condensed and edited for clarity.
References
- Watson, J.L. et al. De novo design of protein structure and function with RFdiffusion. Nature 620, 1089–1100 (2023).
- Bennett, N.R. et al. Atomically accurate de novo design of single-domain antibodies. (2024). Preprint at: https://www.biorxiv.org/content/10.1101/2024.03.14.585103v1