Biology is becoming an information science. The foundation models are already being built — and they need training data that doesn't yet exist.

Perturb is generating large-scale, multimodal perturbation datasets in physiologically relevant human cell systems, designed from first principles for AI training.