Nomadic Emerges with $8.4M to Build Physical AI Training Platform for Robotics and AV Teams

Nomadic, a startup building the spatial intelligence layer for physical AI, today announced $8.4M in funding led by TQ Ventures, with participation from Pear VC, Jeff Dean, and angels & executives from OpenAI, Google DeepMind. Nomadic acts as the visual data engine for robotics and autonomous vehicle teams, enabling continuous monitoring of real-world systems while simultaneously generating production-ready training data and edge-case libraries. The platform is already used by organizations including Zoox, Mitsubishi Electric (Automotive America), Zendar, and others.

Physical AI’s Bottleneck: Video That Doesn’t Turn into Learning Fast Enough

Robotics and autonomy teams collect massive volumes of driving and operational footage. But most of it remains underutilized: reviewed manually, inconsistently labeled, or stored as static archives. As robots move from research environments into large-scale real-world deployment, the bottleneck is no longer model architecture or compute. The bottleneck is understanding what is actually happening across fleets, and turning that understanding into continuous learning.

Teams face two parallel challenges:

• Monitoring: What is happening across our fleet right now? Which failures, regressions, or safety-critical events are emerging?
  
  
• Training: Which data should we use next to improve our models? How can we find rare failures buried inside tens of thousands of hours of footage?
  
  

“Most fleet data goes unreviewed because no human team can get through it all, yet it’s precisely the rare, edge-case footage buried in those archives that matters most for training,” said Antonio Puglielli, VP of Engineering Zendar. “Nomadic makes the full dataset usable, compressing weeks of manual review into minutes so engineers can focus on improving models rather than hunting for the right clips.”

Nomadic’s Solution: Monitoring and Structured Training Data in One Platform
Nomadic is the video intelligence layer that turns raw robotics and AV footage into training-ready data. Its platform analyzes thousands of videos simultaneously to automatically surface failures and key moments, then convert them into structured, searchable events. Instead of treating video as a pile of files, Nomadic transforms it into a living dataset, helping teams validate perception performance, improve data quality, and prioritize what to train next. The result is a faster loop from real-world footage to production models, without needing hundreds of human data labelers.

Key features include:

• Automated event detection that flags critical motion moments without manual review.
  
  
• Compliance analysis to detect potential operational violations and safety issues.
  
  
• AI-powered insights that add analysis and recommendations for detected events.
  
  
• Video search to find similar events across libraries and identify patterns.
  
  
• Natural-language analysis to detect custom scenarios

Multi-sensor uploads so a single run can include front/left/right camera sets as well as LiDAR/logs.

“Teams are sitting on a goldmine of video and sensor data, but most of it never becomes usable training signal,” said Mustafa Bal, cofounder and CEO of Nomadic. “We built Nomadic to bring world-class perception workflows to every robotics team, not just the largest AV labs.”

“Physical AI is going to be won by the teams that can learn fastest from the real world,” said Andrew Marks, Co-founding Partner of TQ Ventures. “Nomadic gives robotics and AV builders the most actionable way to understand their data and rapidly improve their systems.”

Worldclass Founding Team
Nomadic was founded by Harvard computer science graduates Mustafa Bal (CEO) and Varun Krishnan (CTO), who combined their expertise in large scale optimization of distributed systems to tackle the data bottleneck in physical AI. Bal previously worked on AI/ML teams at Snowflake and Microsoft, where he was a core contributor to Microsoft’s DeepSpeed library. Krishnan is a U.S. Chess Master and former INFORMS Wagner Prize finalist for his research on optimizing large-scale systems.

They’ve built a technical team hailing from industry giants like Amazon, Snowflake, and IBM Research, bringing world-class specialization in computer vision, large-scale AI optimization, and production machine learning to the robotics frontier

Companies can try Nomadic today at https://www.nomadicai.com/.

About Nomadic
Nomadic is a physical AI training platform for robotics and autonomous systems. The company converts large video archives into structured, training-ready datasets that help teams monitor real-world performance and accelerate continuous learning. Headquartered in San Francisco, Nomadic works with customers across autonomy, robotics, and industrial applications. Its platform is already used by teams including Zoox, Mitsubishi Electric (Automotive America), Zendar, and others. Nomadic has raised $8.4M from investors including TQ Ventures, Jeff Dean, Pear VC, BAG Ventures, Predictive VC, and Scott Wu. Learn more at https://www.nomadicai.com/.

Contact Information

Riley Munks

PR Advisor

Activate PR - https://www.activate-pr.com/

riley@activate-pr.com