Sensei Robotics

Sensei Robotics is a Boston-based start-up that collects human-demonstration training data for robot learning. It was founded in 2024 and went through Y Combinator's Summer 2024 batch, where it was listed under the name Sensei with a team of two. Its business has two parts: a low-cost hardware platform for recording how people perform manipulation tasks, and a software marketplace that connects robotics companies that need training data with a network of paid human operators, whom the company calls Senseis.

The hardware is a sensorized exoskeleton arm that the operator wears. According to the company's Y Combinator launch post from August 2024, the arm is designed to follow the natural range of motion of a human arm and combines angle, vision and inertial sensors to capture visuo-spatial state information while the wearer works. The company describes the device as a research prototype and says the platform costs less than 300 US dollars, a price point intended to let it equip a large, distributed pool of operators rather than a small number of expensive teleoperation rigs. The examples of tasks given in the launch material are household and light logistics manipulation such as folding clothes, sorting items into bins and loading a dishwasher.

On the software side, a robotics team submits a description of the data it needs. Sensei contractors receive these task descriptions, record demonstrations that fulfil them in their own everyday environments and are paid per contribution, so that the client receives in-the-wild demonstration data for training imitation-learning and other robot AI policies without running its own collection operation. The company's website, senseirobotics.com, carries the tagline Robotic training data, at scale, a Request Data form for robotics companies and a Become a Sensei sign-up for operators joining its beta programme.

The founders both came from Aurora Flight Sciences, the Boeing subsidiary. Chief executive Anubhav Guha left an MIT robotics PhD programme to start the company and was previously chief engineer and programme manager on the Aurora and Boeing team in DARPA's AlphaDogfight trials. Chief technology officer John Piotti graduated from MIT in 2019, worked at Aurora as a robotics engineer leading reinforcement-learning work and earlier as a full-stack web developer at UtilityAPI. Besides Y Combinator, a pre-seed round was recorded in September 2024; its amount and investors have not been disclosed. No customers have been named publicly.