About Heather
Heather Carroll is the Chief Revenue Officer at Path Robotics. Before joining Path, she spent 20+ years in AI, automation, and robotics for supply chain and logistics (warehouses). She was recruited to Path to help scale the business and saw an immediate connection: her husband owns a manufacturing facility in Nashville that struggled with the exact welding challenges Path solves.
About the Company
Path Robotics is building physical AI for manufacturing, starting with welding. Founded by brothers Andy Lonsberry (CEO, PhD in humanoid robotics) and Alex Lonsberry (CTO, PhD in computational neural networks), Path was born from the brothers’ experience growing up in Ohio manufacturing. Andy learned to weld at age 8 in his dad’s garage fab shop. The company has raised $300M through its Series D and employs nearly 200 people in Columbus, Ohio.
Path has ~100 robots deployed in the field and has collected 10+ million inches of weld data which underpins the foundation of their proprietary AI model, Obsidian. Unlike traditional pre-programmed welding robots that fail in high-mix, low-volume environments, Path’s AI uses vision sensors to scan parts, adapt to real-world variation (fit-up gaps, distortion, misalignment), and create fill plans in real-time. The system learns via reinforcement learning from every weld, and can now simulate millions of inches of practice per hour.
Path operates on a Robots-as-a-Service (RaaS) model, guaranteeing production rates and staying embedded with customers to continuously improve performance. Core markets include energy/infrastructure (utility poles, grid components), AI infrastructure (data centers, backup power), and defense—with shipyard partnerships to be announced in early 2025.
Key Topics of Conversation
* Origin Story: Two brothers with PhDs in AI who grew up welding in Ohio, dreaming of an American manufacturing revival
* Why Traditional Welding Robots Fail: Pre-programmed systems can’t handle high-mix, low-volume or imperfect parts, resulting in “robots collecting dust in corners”
* The AI Difference: Vision sensors + real-time adaptation + reinforcement learning from 10M+ inches of weld data
* America’s Welder Crisis: 22% of welders are 55+, another 22% are 45-55; we’re heading for a cliff in 10 years
* Customer Vignette: “$65 million in orders I can’t fill because I can’t find the welders”
* Obsidian Foundation Model: Path’s proprietary weld intelligence trained on real-world data, not physics simulators
* Robots-as-a-Service Model: Why continuous improvement beats one-time CapEx, “like your iPhone camera updating twice a year”
* Shipbuilding Entry: Natural fit for heavy multi-pass welds
* NAVSEA & WPS Compliance: Robot captures every millisecond of weld parameters
* What’s Next: Epsilon model for dexterous manipulation and fit-up; working toward mobile/humanoid embodiments
* Congressional Testimony: Reshoring manufacturing, incentive alignment for RaaS models, federal vs. state AI regulation, permitting bottlenecks
* Data Centers Competing for Welders: $1.2T infrastructure bill + $1T+ private data center investment = same labor pool as shipbuilding
* Lightning Round: Cobots are overhyped; shipyard announcements coming; study welding
Key Links
Follow Heather: https://www.linkedin.com/in/heatherhcarroll/
Follow Austin: https://www.linkedin.com/in/austinelliottgray/
Follow Tim: https://www.linkedin.com/in/tglinatsis/
Subscribe to Rebuilding the Fleet:
* YouTube: https://www.youtube.com/@RebuildingTheFleet
This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit austinegray.substack.com