Vivodyne grows 20 types of human tissue to give AI the data it lacks

For this startup, AI will not cure cancer as long as it learns from static snapshots. Its robotic laboratories produce causal data on living human tissue.

Founded in 2021 by Andrei Georgescu, Doctor of Bioengineering at the University of Pennsylvania, Vivodyne defends a thesis against the current of the ambient discourse on the discovery of IA-assisted drugs: the bottleneck is not the power of models, but the nature of the data on which they train. Available data sets describe frozen states, without ever explaining how a cell moves from inflammation to pathological state.

The company's response lies in HIVE, a modular robotic laboratory capable of cultivating twenty types of human tissues and then measuring and monitoring them independently. The claimed flow is double that of all animal tests conducted in the United States. On toxicity prediction, the company announces 94% accuracy for the liver, 96% for the respiratory tract and 100% for the bone marrow.

Vivodyne raised less than $80 million on two towers led by Khosla Ventures, a modest amount in terms of the value added of the sector. The stated objective is not to sell a model, but the causal raw material that therapeutic combinations need to be evaluated before reaching humans.