Alex Zhavoronkov believes we’re cusp of what he calls “pharmaceutical superintelligence.” Courtesy of Insilico Medicine
Alex Zhavoronkov, featured on this year’s A.I. Power Index, has spent his career pushing the boundaries of what A.I. can achieve in medicine. As founder and CEO of Insilico Medicine, he built Pharma.AI, a platform designed to compress traditional drug development timelines from years to mere months. Already, the system has produced a breakthrough Parkinson’s therapy and delivered promising results in idiopathic pulmonary fibrosis, evidence, Zhavoronkov says, that A.I. is a driver of real clinical progress.
Zhavoronkov believes the industry is on the cusp of what he calls “pharmaceutical superintelligence”: an era when A.I. begins to manage experiments, make decisions and design therapies. That vision is taking shape across Insilico’s fast-growing global footprint. In June, the company raised a $123 million series E round, entered clinical trials for an A.I.-designed cancer drug, launched its Nach01 chemistry foundation model on AWS and expanded its R&D presence in the UAE. Zhavoronkov shares his perspective on misconceptions in A.I. drug discovery, breakthrough clinical moments and where the next wave of innovation will emerge.
What’s one assumption about A.I. that you think is dead wrong?
An assumption about A.I. in drug discovery that I think is dead wrong is the idea that generative models can be trusted without validation. LLM output in biomedicine should clearly not be assumed accurate and needs to be coupled with rigorous experimental validation in the lab, assets obviously need to undergo extensive clinical testing.
If you had to pick one moment in the last year when you thought “Oh shit, this changes everything” about A.I., what was it?
The “oh shit, this changes everything” moment for me was seeing the positive Phase 2a data from our lead asset, rentosertib. We saw signs of potential lung function restoration and improved Forced Vital Capacity (FVC) in patients with Idiopathic Pulmonary Fibrosis. That moment proved to me that A.I. was helping drive real clinical breakthroughs that could directly improve patients’ lives.
What’s something about A.I. development that keeps you up at night that most people aren’t talking about?
How quickly we’re moving toward A.I. training A.I. We’re heading into an era of pharmaceutical superintelligence, where agents won’t just streamline workflows but actually make decisions and design experiments. Most people aren’t talking about it yet, but once A.I. starts managing A.I., everything changes.
How did your A.I. systems actually design the Parkinson’s therapy you announced in August, and what makes this approach fundamentally different from traditional drug discovery methods?
ISM8969 was designed using Insilico’s Pharma.AI platform, which integrates target discovery, molecular generation, and optimization across biology, chemistry and pharmacology. The system first identified NLRP3 as a key regulator of neuroinflammation and then generated structures for oral, brain-penetrant inhibitors. In traditional approaches, a process like this would take years, but our systems compressed it on average to just 12-18 months. We rapidly iterate, test and synthesize 60-200 molecules on average, and our Pharma.AI system produces a candidate for further testing. This candidate showed favorable pharmacokinetics and safety but also delivered dose-dependent improvements in motor function in Parkinson’s mouse models, with effects at the highest dose approaching healthy controls. We streamline the entire process as opposed to traditional drug discovery, which relies on (much more expensive) long trial-and-error cycles. Additionally, a recent study titled “Molecular LEGION: Latent
Enumeration, Generation, Integration, Optimization and Navigation. A case study of incalculably large chemical space coverage around the NLRP3 target” highlighted how we used our Chemsitry42 system and advanced cheminformatics to release over 100 million molecular structures for the NLRP3 target. We uncovered novel scaffolds and patentable chemotypes at a size that traditional libraries would never reach.
Your Life Star 2 lab in Shanghai will be completely A.I. and automated. What specific bottlenecks in drug development are you solving with full automation, and how do you see humanoids changing the research process?
Our new Life Star 2 lab in Shanghai is a leap forward in drug development by being able to fully integrate A.I. decision-making and solve bottlenecks like manual experimental workflows, human bias and fragmented data loops. Our systems can propose targets and orchestrate workflows, while our modules execute cell culture, high-throughput screening, next-generation sequencing, cell imaging and genomics analysis and prediction, without human intervention. The automated lab is faster and more precise than humans, and as they perform experiments, they feed the A.I. system with data, improving the system’s target hypotheses and ability to validate those hypotheses. Our bipedal humanoid “Supervisor” allows fully automated operation on lab equipment originally designed for humans and can handle tasks like pipetting, reagent handling, and real-time lab oversight. As training continues, so will the elevation of tasks.
You’ve moved from longevity research into broader disease applications like cancer and lung disease. How does your A.I. platform adapt across different therapeutic areas, and where do you see the biggest opportunities for A.I. to accelerate clinical breakthroughs in the next 3-5 years?
We’ve always pursued diseases that were closely linked to aging processes. Our first internally developed program targets idiopathic pulmonary fibrosis, and our algorithms used in longevity research can be trained just as effectively on oncology, fibrosis or neurodegeneration. PandaOmics uncovers novel targets from large, complex biomedical data, and Chemistry42 generates novel molecules against those targets. In the next couple of years, I see a great opportunity in further shortening the time between a new target and proof of concept in the clinic. In areas of high unmet need, A.I. can give us the ability to move cheaply, more efficiently and with potentially greater precision than traditional approaches. I believe these systems will evolve into what I call pharmaceutical superintelligence, where A.I. supports discovery and actively drives decision-making across the entire drug development process.
