Maxinsights raises $30M seed to turn everyday chores into robot training data

What Maxinsights does

Maxinsights converts video footage of everyday tasks—such as cooking, cleaning, and manual work—into training data for robots. The company pays contractors in six countries to wear cameras while performing these activities, creating a dataset of over 2 million hours of footage. Of this, 1.5 million hours are already prepared for use in AI models.

The startup’s approach addresses a critical gap in robotics: the lack of large-scale, real-world datasets akin to the text corpora that powered language models. As John Koetsier of Forbes noted, while language models learned from decades of internet text, robots have no comparable resource. Maxinsights aims to fill this void by generating data that captures the nuances of human movement and interaction with objects.

New products and efficiency gains

Alongside the funding announcement, Maxinsights launched two products: MaxLoop and MaxGlove. MaxLoop is software that automatically reviews incoming footage, checking for quality, task completion, and proper camera placement. According to Maxinsights, the tool has reduced review costs by 80% and cut feedback time from weeks to the same day.

MaxGlove, developed in collaboration with Genesis AI, is a sensor glove that records grip force, object weight, and how contact shifts during tasks. The glove aims to solve a key challenge in robotic data collection—capturing the tactile details of human interactions. Maxinsights claims existing alternatives require lengthy calibration and yield only 10 to 20 minutes of usable data per day, while MaxGlove promises significantly higher efficiency.

Customer traction and market expansion

Maxinsights has grown its customer base to over 30, including frontier AI labs, with eight clients on recurring contracts. Notable customers include Genesis AI, for which Maxinsights supplied 200,000 hours of first-person data for the GENE-26.5 model, and Dyna Robotics, where it is the largest data provider for the Dyna-2 model.

The startup processes about 450,000 hours of footage per month, reflecting its scalable infrastructure. To sustain this growth, Maxinsights plans to expand its operations to Turkey and Mexico, building on its existing presence in China and Indonesia.

Founders and funding context

The $30 million seed round was led by Peak XV, with participation from Tola Capital, South Park Commons, INCE Capital, and Alumni Ventures. Rohit Agarwal, Managing Director at Peak XV, emphasized the potential of Maxinsights’ data-centric approach to accelerate the development of physical AI.

Maxinsights was founded by Yuguang Yuan, who previously co-founded Mega Gear Robotics, and Kun Wang, an early engineer at Google X and Waymo, and co-founder of QCraft, a self-driving company. The founders’ backgrounds in robotics and AI infrastructure underscore the startup’s focus on bridging the gap between human tasks and robotic learning.

The funding round reflects broader investor enthusiasm for physical AI. According to The Next Web, investors poured $47.4 billion into the sector in the first half of 2026, nearly four times the amount invested in the second half of 2025.

Market context for robotics training data

The surge in funding for Maxinsights aligns with a broader trend of investment in physical AI and robotics. Earlier this year, competitors like Encord and Enigma raised €50 million and $70 million, respectively, for their own data platforms and robotics initiatives. Figure, another player in the space, runs a similar program called Index, where it pays users to film chores, though it focuses exclusively on serving its own models.

This trend mirrors developments in other sectors, such as autonomous vehicles, where companies like Waymo have secured significant funding to scale their datasets and testing infrastructure. The regulatory exemption granted to Aurora Innovation and Kodiak AI for self-driving trucks also highlights how data collection and validation are becoming central to the adoption of robotics and AI-driven systems.

For AI startups, the challenge of sourcing high-quality training data is a recurring theme. Maxinsights’ model—paying contractors to generate real-world footage—offers a scalable alternative to synthetic data or simulated environments, which often lack the complexity of human interactions.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

Maxinsights’ funding and product launch signal growing recognition of training data as a bottleneck—and an opportunity—in robotics. For founders and operators in AI and robotics, the startup’s model demonstrates how real-world data collection can be scaled efficiently, even in a fragmented market like household tasks. The $30 million seed round, led by a prominent investor like Peak XV, also reinforces the importance of infrastructure plays in AI, particularly those that address gaps in data quality or accessibility.

For startups building robots or AI models, Maxinsights’ customer roster—including frontier labs—shows that outsourcing data collection is becoming a viable strategy. The launch of MaxLoop and MaxGlove further highlights how automation and hardware innovations can reduce the cost and complexity of generating training data. Competitors like Figure, which rely on internal datasets, may face pressure to match the scale and diversity of Maxinsights’ offerings.

Key takeaways

• Maxinsights’ $30M seed round underscores the growing investor interest in scalable solutions for robotics training data.
• The startup’s ability to process 450,000 hours of footage per month and reduce review costs by 80% demonstrates operational efficiency in a labor-intensive space.
• Expansion into Turkey and Mexico suggests a strategy to diversify data sources across geographies and task types, which could improve model generalization.
• For AI and robotics startups, Maxinsights’ model of outsourcing data collection offers a potential alternative to synthetic data or in-house programs.

Related on Lazyfounder

• Peak XV
• Genesis AI
• Figure
• More ai news