Maxinsights Raises $30M to Improve Physical AI Data

The big picture: Maxinsights has raised $30 million in seed funding and launched two new products aimed at enhancing the collection and processing of real-world training data for physical AI systems.

Why it matters:

• Real-world complexity: Physical AI systems, such as robots, operate in unpredictable environments, requiring vast amounts of ‘messy’ real-world data rather than staged scenarios to handle edge cases effectively.
• Scalable quality: Maxinsights has developed an end-to-end pipeline to capture and maintain high-quality data across millions of hours from thousands of contractors, addressing a critical bottleneck in physical AI development.
• Industry adoption: The company is a key data provider for prominent AI models, including Genesis AI’s GENE-26.5 and Dyna Robotics’ Dyna-2 World-Action Model, underscoring its impact on advanced robotics and embodied AI.

How it works:

• Integrated pipeline: Maxinsights manages the entire data lifecycle in-house, from collection using proprietary camera hardware and a network of over 5,000 contractors, through quality control and annotation, to curation and delivery.
• Automated and human review: Footage undergoes initial checks for quality and task completion, followed by proprietary models that label environmental details, tasks, subtasks, instructions, hand poses, and object interactions.
• New tools: The company introduced MaxLoop, software for real-time data collection evaluation and correction, and MaxGlove, a sensor glove developed with Genesis AI to capture crucial touch and force information for more nuanced robot interactions.

The catch: The demand for high-quality physical AI training data is intensifying, attracting more players and increasing competition. Maxinsights’ challenge will be to scale its global contractor network and maintain data quality across diverse environments while fending off new entrants. The development of specialized hardware like MaxGlove highlights the increasing complexity and cost associated with capturing the nuanced data required for truly robust physical AI.

Key Facts

• Company: Maxinsights
• Amount: $30M
• Round: Seed
• Investors: Peak XV (lead), Tola Capital, South Park Commons, INCE Capital, Alumni Ventures
• Founders: Yuguang (YG) Yuan, Kun Wang
• Announced: 2026-10-23
• Sector: Physical AI Training Data
• Headquarters: Santa Clara, CA, US