Peak XV leads $30 million seed in robot-data start-up Maxinsights

Maxinsights, a start-up from Santa Clara, California, that pays thousands of people to wear cameras while they cook, clean and do other everyday work, raised $30 million in a seed round led by Peak XV, the company announced on October 22, 2026. The footage is sold as training data to companies building artificial intelligence (AI) for robots. Tola Capital, South Park Commons, INCE Capital and Alumni Ventures also took part. Alongside the funding, the company launched two products: MaxLoop, a data-quality tool, and MaxGlove, a sensor glove that records touch and grip.

The AI systems behind chatbots learned from huge amounts of text on the internet. Robots have no such library. To learn how to fold a towel, load a dishwasher or chop vegetables, a robot model needs many hours of examples of people doing exactly that, ideally seen from the person's own point of view. This is often called "physical AI": AI that acts in the real world, not only on a screen. Shortage of good training data is one of the main things holding it back.

Maxinsights is betting that the fastest way to fill that gap is to pay ordinary people to record their normal work. Other companies rely more on computer simulations or on staged sessions in labs. The company argues that real homes and workplaces, with their mess and variety, teach robots more. Its claim that it supplied the first-person data behind a model from Genesis AI, a robotics AI developer, is the company's own statement; outside observers note that the real test is whether such data makes robots perform better outside controlled settings.

The new glove points to the next step. Video shows what a hand does, but not how hard it grips or how heavy an object is. Adding touch and force data could help robots handle fragile or heavy items, which is where many fail today.

For Indian readers there are two threads. The round, led by Peak XV, shows venture money moving into the data layer that robotics will need worldwide, an area where Indian start-ups could also compete. And the business model, paying a large distributed workforce to create AI training data, is similar to the data-labelling work that already employs many people in India. Maxinsights has not said which six countries its contractors are in; reports mention offices in China and Indonesia and plans for Turkey and Mexico.

Short-term: The company plans to grow its contractor network and data volume and to roll out MaxLoop and MaxGlove to customers. Rivals in robot training data will face pressure to show similar scale.

Long-term: If real-world video becomes the standard way to train household and service robots, a new kind of gig work, recording daily tasks on camera, could grow across many countries. That will raise questions about pay, privacy for people who appear in the footage, and who owns the recordings.

Who is affected: Robotics and AI companies that need training data; gig workers who may be paid to record their work; investors in the fast-growing physical AI sector; and, eventually, households and businesses that may use the robots trained on this data.

Questions and Answers