The elusive goal of automating physical scientific work just got a $25.0M boost. Transfyr, a startup building "physical AI for science," has closed a Seed round to tackle the missing record of experimental execution that plagues research, making it hard to learn from failures or transfer know-how.
The company aims to provide scientists with the equivalent of an athlete's instant replay, capturing the small decisions and invisible actions that determine experimental success or failure. By turning real scientific work into a high-fidelity, machine-readable record, Transfyr is also building a large commercial dataset on real-world scientific execution, providing crucial infrastructure for training and model development.















