High-fidelity robot world models that enable robotics companies to evaluate and train robot policies in simulation before physical deployment.
Fern provides RL environments powered by a foundational world model that robotics companies use to run and evaluate robot policies in software across varying production environment scenes. It generates RL training heads on top of existing policies to improve speed and success rate, creating a continual learning flywheel where production robots improve weekly. The platform enables plug-and-play testing of robot policies completely in software before physical deployment. Fern is a product of Fern.