Duet Labs
Duet Labs is building a new data layer for physical AI by shipping multimodal robotics hardware and data infrastructure for frontier research labs. The company combines sensing hardware with collection and processing pipelines to feed AI models the real-world training data they need. Their work includes published ML research through NeurIPS, IEEE, IIAI, and Harvard, informing their technical approach.
- Robotics
- Artificial Intelligence
- Hardware
Funding
Founders
Founder details are not available yet.
Product
Problem
Physical AI systems—robots and embodied agents—require vast amounts of high-quality multimodal data to learn and operate effectively. Current data infrastructure is fragmented and inadequate, limiting the ability of robotics companies to scale their models and achieve reliable real-world performance.
Solution
Duet Labs is constructing a new data layer specifically designed for physical AI, combining multimodal sensing hardware with data infrastructure to support robotics development. The company ships robotics hardware tailored for frontier lab contracts, providing the physical instrumentation needed to capture rich, real-world data. Their platform integrates sensing, collection, and processing pipelines to feed AI models with the training and operational data they require. By building this foundational data infrastructure, Duet Labs aims to unlock the next scaling law in robotics—where data volume and quality become the primary drivers of model capability.
Target Audience
Primary customers are frontier AI research labs, advanced robotics companies, and organizations developing physical AI systems that require high-quality multimodal training data at scale.
Features
- Multimodal robotics hardware designed for collecting diverse sensor data in real-world environments
- Data infrastructure tailored to physical AI training and deployment workflows
- Hardware solutions built explicitly for frontier research lab contracts and advanced robotics use cases
- ML research capabilities with publications through NeurIPS, IEEE, IIAI, and Harvard, informing technical design choices