Vivodyne Launches Innovative AI-Driven Lab to Address Data Gaps in Drug Discovery

Vivodyne, a biotech startup emerging from the University of Pennsylvania, has unveiled a solution to what it describes as a critical data problem in the AI drug-discovery industry. The company’s modular robotic labs, known as HIVE, can cultivate 20 types of human tissue, autonomously administer treatments, and monitor biological responses. This new approach aims to generate causal biological data that current AI models typically lack, which is often derived from animal testing and isolated studies of cells or proteins.

Andrei Georgescu, the co-founder and CEO of Vivodyne, emphasized the shortcomings of existing AI models, stating, “Absent human testing, what are these [AI] models going to do? They’re going to cure cancer in mice.” This skepticism echoes sentiments expressed by Dario Amodei, CEO of Anthropic, who noted that grand claims about AI curing cancer have become more cliché than credible.

Despite past predictions from tech leaders, including Sam Altman of OpenAI and Demis Hassabis of Google DeepMind, asserting that AI could dramatically reshape healthcare, the results have been mixed. A few AI-designed drugs have progressed to human trials, including one that reached Phase III, but significant hurdles remain that are not necessarily solvable by current AI capabilities.

The field has seen some advancements, such as AlphaFold, which has deepened the understanding of protein structures but has yet to lead to a new drug. Isomorphic Labs, a company established to further pursue drug discovery using AlphaFold, has pushed its first trials from 2025 to later this year. In its communications, Isomorphic Labs has emphasized the necessity for more precise predictive models covering a wide range of biochemical properties and interactions.

Georgescu pointed out the urgent need for a “sanity check” in the industry, noting that many existing models fail to adequately capture the complexities of human biology. This inadequacy contributes to a significant challenge in the pharmaceutical sector, where approximately 90% of drugs that show effectiveness in animal testing do not gain approval in clinical trials for human use.

Vivodyne, which has attracted nearly $80 million in funding through two rounds led by Khosla Ventures, recently established what it claims is the world’s largest “human data center” just outside San Francisco. Georgescu reported that his team is currently achieving double the throughput compared to all animal trials conducted in the U.S.

The company’s innovative approach aims to streamline the drug development process by providing more reliable insights before costly clinical trials, which can average tens of millions of dollars. While Vivodyne is collaborating with several major pharmaceutical companies, it remains tight-lipped about its partners. Georgescu likened the endeavor to automotive safety testing, where manufacturers traditionally ensure confidence in their vehicles passing regulatory checks before actual testing, a confidence lacking in drugmaker trials leading to FDA approval.

Furthermore, Georgescu envisions these autonomous biology labs as pivotal in producing the causal data required to train advanced models focused on human biology. He cited recent studies revealing that generative AI training is often based on static representations of cells, lacking understanding of how those cells reached their respective states. He stated, “All the training is done on static snapshots of these cells, and the models are not conditioned at all by how a cell got to that state.”

The HIVE machines at Vivodyne are designed to monitor thousands of dynamic experiments, where diseased tissue is subjected to various stimuli. Georgescu anticipates that this reinforcement learning approach will yield AI models that can make significant contributions to healthcare advancements. He believes this capability will be essential not only for addressing current medical challenges but also for developing therapies targeting multiple biological pathways in complex diseases.

“If we want combination therapies, the space that has to be searched explodes—it can’t be an experimental approach,” he asserted. “You have to say, ‘I want this effect to happen, so what cause should I invoke?’ Establishing causality in human biology is the basis of all of this.”