Skip to main content
F

FortyFive

FortyFive builds cloud-based simulation infrastructure for robotics, with the DreamLake platform for synthetic data generation, annotation, sim-to-real transfer, and realistic evaluation of robotic systems.

FortyFive (FortyFive Labs) builds simulation infrastructure for robotics. The company spun out of MIT CSAIL and has pioneered techniques for training robots using only synthetic data - an approach that sidesteps much of the real-world data collection bottleneck in robot learning. Its work sits at the intersection of robotics simulation and generative AI, producing synthetic training data and realistic evaluation environments for robotic systems.

The company's main product is DreamLake, a cloud-based platform covering simulation, data generation, annotation, and evaluation. DreamLake is intended to help robotics teams move to AI-native development workflows: generating diverse training data at scale, narrowing the sim-to-real gap, and providing environments in which robotic systems can be evaluated realistically rather than against ad-hoc test cases. The toolchain addresses both sides of the robotics ML pipeline - producing and labeling the data that physical AI models are trained on, and standing up the simulation infrastructure to validate them.

FortyFive is a lean team of researchers and engineers drawn from DeepMind, OpenAI, FAIR, MIT, Berkeley, Caltech, Harvard, and Yale, working across the United States and China. The group operates as a research-oriented outfit bridging academia and industry, with a technical focus spanning robotics simulation, synthetic data generation, generative AI, data annotation, sim-to-real transfer, and evaluation environments for robotic systems.

1 Open job

We use cookies

We use cookies to understand how this board is used and to improve it. Analytics run only if you accept. Cookie Policy