02 March 2026 | Interaction | By Editor Robotics Business NEWS <editor@rbnpress.com>
In this exclusive interview with Robotics Business News, Andy Lonsberry, CEO and co-founder of Path Robotics, shares how physical AI is redefining industrial automation through intelligent robotic welding systems. Drawing on tens of millions of welded inches of real-world data, Path Robotics’ Obsidian model enables robots to perceive, reason and adapt in real time — unlocking new levels of productivity, quality and flexibility in shipbuilding and heavy manufacturing while helping address growing skilled labor shortages.
Path Robotics has trained its physical AI on tens of millions of welded inches — how does this data advantage translate into real-world performance improvements in complex shipbuilding environments?
Obsidian is Path Robotic’s foundational physical AI model for welding. Trained on tens of millions of welded inches, Obsidian transforms a traditional industrial robot arm from a rigid, repeat-only machine into a real-time perception and decision-making system that can see, understand, and adapt in complex shipbuilding environments.
Welding in shipbuilding is often performed on massive assemblies where fit-up can vary significantly. Path Robotics addresses this challenge through its proprietary vision system, which scans every welding seam using proprietary lasers and optics. This data is then fed to Obsidian to develop the weld and path plan in real time. Once welding, the system monitors the weld pool and seam to compensate for any heat distortion as the weld progresses. The result is consistent weld quality without the traditional burden of fixturing parts to perfection or reprogramming every time the part changes.
Path Robotic's Intelligent Welding Cells adapt in real-time for an agile flow of high-quality welds – designed for these types of high-mix, high-variability environments. Each of our welding cells adds real production data, improving the model and delivering updates that raise quality and speed across the board. In one customer example, we were able to reduce manual welding time per chassis by 91% – from 150 hours to 13 hours.
Shipbuilding presents highly variable welding conditions compared to factory manufacturing. How does Path’s physical AI enable robots to see, reason, and adapt in such unpredictable environments?
Traditional automation in shipbuilding can only follow a predefined path – the robots follow and weld that same path every time, regardless of what's in front of them, until the robot is changed, a tedious process involving a scarce, but highly skilled robotics programmer. That works fine if all parts are identical. Or if spending a few months programming a robot, designing and producing perfect fixturing is worth it because it will pay for itself over millions of parts.
But the reality of most shipyard manufacturing is messy. Upstream parts come in slightly differently each time. Fit-up varies, joints wander, and heat distorts the metal as you weld. A skilled welder sees all of that and adjusts on the fly. Traditional automation can't. So, despite more than a century of welding automation, 80% of welding is still done manually. The technology to perceive what's actually there and adapt in real time just didn't exist until now.
Path Robotics’s AI-driven autonomous welding technology presents an opportunity for leaders like Saronic to expand distributed shipbuilding capacity and augment the shipbuilding workforce. The ability to see, reason, and adapt is what makes Path Robotic’s physical AI incredibly viable in shipyards, where variability is the norm rather than the exception.
This collaboration marks Path Robotics’ expansion into maritime manufacturing. What unique technical or operational challenges did you anticipate when entering the shipbuilding sector?
America’s shipyards are being called on to scale at an unprecedented pace, intensifying the need for highly skilled welders. Physical AI offers a powerful way to meet that moment, pairing the expertise of American shipbuilders with intelligent automation.
By integrating real-time decision-making directly into robotic welding systems, we’re helping shipyards adapt to the complex, ever-changing reality of most manufacturing environments – unlocking capacity that traditional automation can’t. The result is a new era of advanced manufacturing that strengthens the workforce and expands what’s possible on the shipyard floor.
Skilled labor shortages remain a major concern across heavy industries. How do you see physical AI augmenting — rather than replacing — experienced welders in modern shipyards?
The “Golden Fleet” is the U.S. Navy’s initiative to modernize and build the next generation of battleships. It’s estimated that U.S. naval shipbuilders must hire more than 250,000 people over the next decade to support the construction of that fleet. What’s more, U.S. manufacturing faces a critical shortage of welders in particular. The American Welding Society projects that the U.S. will need more than 320,000 new welding professionals by 2029 – about 80,000 jobs to be filled annually.
This increased demand coupled with the significant workforce gap is pushing the industry to find ways to scale output without adding headcount from a labor pool that simply does not exist.
At the same time, traditional automation fails because most real-world parts are not perfect. Traditional welding robotics excels at consistency, but simply cannot adapt to the messy reality of most manufacturing environments, where things are always changing. Path Robotics is addressing each of these challenges – pioneering physical AI to enable robots to weld with the judgment and adaptability of a veteran welder.
Can you explain how real-time intelligence embedded in robotic welding systems changes traditional automation models used in heavy manufacturing?
In traditional automation, a robot is programmed with a fixed path with the requirement that every part needs to be identical. That process is challenged in heavy manufacturing industries like shipbuilding where variability is constant.
By embedding real-time intelligence in robot welding systems through physical AI, we are empowering that process. Instead of simply executing instructions, the robot is now able to see the environment, reason about what it sees, and adapt in real time. This shifts traditional automation away from merely repeating what is programmed, to now understanding and responding accordingly. In doing so, it enables manufacturers to reduce some of the major bottlenecks we see in shipbuilding and heavy industry – like preprogramming, manual rework, and overreliance on perfect fixturing.
What measurable production gains — whether in quality, throughput, or safety — do you expect shipyards to achieve through intelligent welding cells?
With the implementation of Path Robotic’s physical AI for welding into shipbuilding with leaders like Saronic, the industry could see significant measurable gains, including unlocking capacity to accelerate throughput and improved quality.
Path Robotic’s physical AI enables more consistent welding on-time and reduces the amount of rework and reprogramming required in traditional manufacturing environments. In shipyards, that benefit translates to increased production capacity without requiring proportional increases in labor.
In addition, because Path Robotic’s physical AI is responding to real-time part conditions and variations, we can ensure quality, while also reducing manual inspection failures and downstream rework.
As physical AI adoption grows, what role will software and data feedback loops play in continuously improving welding performance over time?
Every weld completed by Path Robotic’s intelligent welding cells generates real data, including seam geometry, heat input, travel speed, weld pool behavior, and more. That data feeds back into the model, refining its ability to see, reason, and adapt in future welds. This creates a compounding performance curve. Unlike traditional automation, physical AI systems improve as they operate. Updates can be deployed across cells, raising performance fleet-wide. Over time, this creates a data flywheel where each installation strengthens the broader system and creates a self-improving loop.
However, manufacturing has a higher reliability threshold than AI that remains in the digital world. A software demo that works 70% of the time and then crashes might mean restarting your computer. But in a shipyard environment, production systems must operate at greater than 99% reliability – otherwise the implications of a failed weld could mean scrapping a million dollar part and delaying a project by six weeks.
That’s why long-term data accumulation is critical. Years of weld data in real production environments are what allow these systems to move from mere demonstrations to real manufacturing environments. And as adoption scales, those feedback loops will only further accelerate development and push physical AI to being a standard operating model in industrial facilities.
Looking ahead, how do you envision physical AI reshaping U.S. industrial competitiveness and the future of advanced manufacturing beyond shipbuilding?
AI that exists in the digital world has already transformed how companies process information and make decisions, making them more efficient and dramatically accelerating production output. Physical AI extends that transformation into the physical world – from factories to shipyards to other critical infrastructure.
We believe 2026 marks the inflection point for physical AI awareness and adoption. Market understanding is accelerating rapidly, much like digital AI did several years ago. The difference is that physical AI solves a harder problem – enabling machines to perceive and act in complex, dynamic environments with the reliability required for industrial production.
Over the next several years, as these systems move from pilot environments into full-scale production, physical AI will enable: distributed manufacturing capacity without proportional labor expansion, greater resilience in supply chains, faster scaling of defense and critical infrastructure programs, and a new generation of intelligent factories that continuously learn and optimize.
For the U.S., this has strategic implications. Industrial capacity is directly tied to economic strength and national security. By embedding artificial intelligence into manufacturing systems, physical AI can help rebuild domestic production capability in sectors that have historically struggled with labor shortages and variability-driven inefficiencies. Shipbuilding is an early proving ground because it represents one of the most complex, high-stakes manufacturing environments in the world. But the long-term impact extends across heavy industry to energy, transportation, aerospace, and beyond.
Physical AI has the potential to redefine what can be manufactured, how quickly it can be made, and where production can take place, strengthening American industrial competitiveness.
