Maxinsights Raises $30M Seed Funding Led by Peak XV for Robot Training Data

Maxinsights, a Santa Clara, California-based startup building data infrastructure for robotics and physical AI, has raised $30 million in seed funding led by Peak XV. The company plans to expand its human video data collection operations, develop its software and hardware, and support its global business.

The funding round also included Tola Capital, South Park Commons, INCE Capital and Alumni Ventures. Maxinsights announced the financing on October 22, 2026, according to its funding announcement reported by Tech Funding News.

The startup operates a network of more than 5,000 contractors across six countries who record everyday activities, including loading dishwashers, preparing food and washing cars. The resulting first-person videos are processed and supplied to robotics companies and AI laboratories to help train models to perform tasks in real-world environments.

Maxinsights says it has captured more than 2 million hours of footage, with 1.5 million hours ready for model training. It serves more than 30 customers, including frontier AI laboratories, and reports that eight customers have recurring contracts.

Turning Everyday Human Activities Into Robot Training Data

Robots need exposure to a wide range of real-world situations to learn how to interact with objects and complete practical tasks. Activities that appear simple to humans, such as folding clothes, preparing a meal or cleaning a kitchen, can involve numerous movements, changing environments and unexpected obstacles.

Training robots to handle these situations requires large volumes of data showing how people perform tasks in different conditions.

Maxinsights aims to meet this demand by collecting first-person footage of people carrying out everyday work in homes, restaurants and hotels. Workers wear the company's proprietary camera hardware while completing tasks at natural speed.

The footage is then processed and prepared for use in robotics training.

According to the company, its collection approach focuses on real-world environments rather than relying exclusively on staged tasks or controlled demonstrations. The goal is to capture the variations and unexpected situations robots may encounter when deployed outside laboratories.

This data can help robotics developers expose their models to a broader range of human activities and environmental conditions.

A Five-Step Pipeline for Robotics Data

Maxinsights has built an end-to-end system for collecting, processing and delivering training data.

Its process begins with first-person video collection through its contractor network. The company then checks footage for quality, annotates it using proprietary models, curates datasets according to specific tasks or environments, and delivers raw or processed data to customers.

Customer feedback helps guide subsequent data collection, allowing the company to adjust its datasets to meet changing training requirements.

Maxinsights says it developed the pipeline internally, including its hand-tracking models and software for assessing footage as it arrives. The company argues that collecting large volumes of video is only part of the challenge: the data must also be consistent and useful for training.

The startup's founder and CEO, Yuguang (YG) Yuan, said that maintaining quality across millions of hours of footage and thousands of workers is a central part of the company's approach.

By building its own collection and processing infrastructure, Maxinsights aims to provide robotics developers with data that is ready for use rather than requiring them to build their own large-scale collection and quality-control operations.

Quality Control Is a Key Differentiator

First-person footage can be difficult to process automatically. Hands may disappear behind objects, multiple people may enter the same frame, and gloves, jewellery or motion blur can complicate tracking.

Maxinsights says it trained its own hand-tracking models on footage collected through its network and built a platform called MaxLoop to assess clips as they arrive.

The company reports that operators now receive feedback on their footage on the same day, compared with a previous turnaround of one to two weeks. It also says review costs have fallen by approximately 80%, its human review team is 90% smaller, and its system processes 450,000 hours of footage per month.

These figures are company-reported metrics and have not been independently verified in the supplied report.

The focus on quality control could be commercially important because robotics models depend not only on the quantity of training data but also on whether that data accurately represents useful actions and real-world conditions.

Peak XV managing director Rohit Agarwal highlighted this approach when discussing the investment, arguing that Maxinsights treats quality control as an engineering challenge rather than simply a staffing expense.

Customers Include Genesis AI and Dyna Robotics

Maxinsights has identified Genesis AI and Dyna Robotics among its customers.

The company says it supplied all 200,000 hours of first-person footage used to train Genesis AI's GENE-26.5 model. It also describes itself as the largest data provider behind Dyna Robotics' Dyna-2 model.

Genesis AI previously raised $105 million in seed funding in July 2025, according to the report. Dyna Robotics has also discussed the use of human video data in training its robotics model.

These relationships illustrate the potential demand for external data suppliers as robotics developers work to train systems capable of completing increasingly complex physical tasks.

However, robotics companies are also developing their own data collection capabilities. Some are building dedicated training environments or collecting footage internally, which could create competition for third-party providers.

Maxinsights will need to demonstrate that outsourcing data collection and processing offers customers sufficient quality, scale and efficiency to justify continuing commercial contracts.

Physical AI Funding Is Growing

The investment comes amid increased funding for physical AI, a category that includes robotics, autonomous vehicles, drones, industrial automation and related technologies.

According to Crunchbase figures cited by Tech Funding News, physical AI companies attracted $47.4 billion across 521 deals in the first half of 2026, almost four times the $12 billion raised in the second half of 2025. The category includes a broad range of businesses, and the total was significantly influenced by large financing rounds such as Waymo's $16 billion Series D.

The growing investment in robotics and autonomous systems is increasing attention on the infrastructure needed to develop them, including training data, simulation environments, sensors and specialised hardware.

Maxinsights is targeting one part of this ecosystem: the collection and processing of human demonstrations that can help robots learn physical tasks.

Other startups are pursuing similar opportunities. Mecka AI, for example, has also developed a human-data collection business for robotics, while some robot manufacturers are investing in dedicated training facilities to generate their own data.

This creates a market opportunity for specialised suppliers, but also means Maxinsights must compete on data quality, cost, coverage and delivery speed.

Founders and Technology Team

Maxinsights was founded by Yuguang (YG) Yuan, who previously co-founded industrial robotics company Mega Gear Robotics, which was acquired.

According to the company's funding announcement, Yuan also previously held an executive leadership role at Eaton, overseeing product groups with approximately $150 million in annual business and more than 1,500 employees.

Chief technology officer Kun Wang was an early engineer at Google X and Waymo and later co-founded QCraft, an autonomous driving technology company. Maxinsights' broader team includes engineers with experience at companies such as Google, Microsoft, Meta, Waymo and Hulu.

The team's experience spans robotics, autonomous systems, software and large-scale operations, all relevant to building a data collection platform intended to serve AI developers.

How Maxinsights Plans to Use the Funding

The $30 million seed round will support the company's next stage of expansion.

Maxinsights plans to increase its data collection capacity, advance its software and proprietary camera hardware, and strengthen its global operations.

The company is positioning itself as a specialised supplier for robotics developers that need large amounts of real-world human activity data but may not want to build and manage their own collection networks.

Its reported scale, customer base and processing capabilities provide an initial indication of demand. However, the company has not disclosed revenue figures in the supplied report, making it difficult to assess its commercial performance beyond its stated customer and operational metrics.

As robotics systems move from controlled demonstrations toward practical deployment, demand for diverse, high-quality training data could become increasingly important.

Maxinsights is betting that a scalable network of human workers, combined with automated quality control and data processing, can become a valuable part of the infrastructure behind the next generation of robots.