Persona AI: Humanoid Robots for Shipyard Welding

What Persona AI Builds

Persona AI builds industrial humanoid robots aimed at a specific, narrow job: welding inside shipyards and heavy-fabrication shops. That is a deliberately different starting point than the household-chore or warehouse-picking demos most humanoid robotics companies lead with, and it maps to a real labor shortage. Skilled shipyard welders are difficult to hire and train, and the work itself is physically demanding and often unsafe.

The company was co-founded in 2024 by Nic Radford, who previously led NASA's Valkyrie humanoid space-robot program, and Jerry Pratt, formerly the CTO of humanoid maker Figure AI. That pairing of space-robotics and commercial-humanoid backgrounds is a specific bet: build a humanoid narrow enough to be commercially viable in 2026, rather than broad enough to be a general-purpose worker on day one.

Persona's Gen 1 humanoid stands 175 centimeters tall and weighs 85 kilograms, dimensions chosen to match a human welder's physical envelope, so the robot can access the same work positions, use the same scaffolding, and move through the same passageways a human welder already does, without a shipyard redesigning its workflow around the robot.

Funding and Why Welding Specifically Is the Hard Engineering Problem

Persona AI has raised more than $27 million to date, funding that scales a team building for a genuinely harder manipulation problem than most humanoid demos attempt. Welding is not generic pick-and-place work. A weld pass has to track a joint line within a few millimeters at a controlled travel speed and arc distance, in a heat, spark, and fume environment that also has to be managed for the robot's own hardware, not just for a human operator's safety gear.

That is why Persona frames its Gen 1 unit around the physical envelope of a human welder rather than around raw payload or reach numbers the way an industrial-arm spec sheet usually does. A welder's actual working conditions, cramped hull sections, overhead positions, ladders, and scaffolding, are the constraint that decides whether a robot can do the job at all, before weld quality even becomes the question.

The 175-centimeter, 85-kilogram spec is a proxy for a claim the company has not yet had to prove at full production scale: that a robot sized like a person can be walked through a shipyard's existing infrastructure without a facility retrofit. If that claim holds, it removes a cost line that trips up most industrial-robot deployments, since new fixturing and worksite changes routinely cost more than the robot itself.

The HD Hyundai Deal and What It Actually Covers

Persona AI has a signed agreement with HD Hyundai, one of the world's largest shipbuilders, to deploy the welding humanoid inside its yards, alongside a separate steel-fabrication pilot in Louisiana. The stated timeline delivers prototype humanoids by the end of 2026, with field testing and commercial deployment beginning in 2027: a multi-year runway, not an immediate fleet rollout.

That timeline matters for how to read the announcement. A signed deployment agreement with a named shipbuilder is a real commercial validation signal, since HD Hyundai is putting its own yard and welding standards behind the pilot, but it is still a prototype-and-field-test phase, not proof the robot already welds at production rate and quality inside a real hull.

In the near term, Persona's Gen 1 units are guided through teleoperation. A human operator drives the robot through real welding tasks, and that operator-driven data is what trains the robot's AI toward eventual autonomous operation. That is the same category of data-collection strategy other robot-learning efforts use, teleoperated task data first and autonomy after, rather than a robot welding unsupervised from day one.

Why Welding Is a Defensible Near-Term Humanoid Niche

Shipyard welding has three properties that make it a better near-term humanoid target than a general household or warehouse task: the work is done in spaces literally built for a human body (scaffolding, ladders, confined hull sections), the skilled-labor shortage is acute and well documented across the shipbuilding industry, and the task itself, following a weld path with consistent technique, is more repeatable and instrumentable than open-ended chore work.

The company has said it sees near-term economic viability specifically in heavy industrial humanoid deployment, not in racing toward a general-purpose home robot. That framing is consistent with the HD Hyundai deal: a narrow, high-value, labor-constrained task with a willing enterprise customer beats a broad consumer promise with no paying deployment yet.

The tradeoff is scope. A humanoid tuned for shipyard welding does not automatically transfer to a different industrial task without new teleoperated training data, and Persona AI's own stated roadmap, a 'multipurpose future' beyond welding, is explicitly a later-stage ambition, not a current capability.

• Confirm whether a cited deployment is a signed pilot/agreement or a live production welding operation.
• Ask what share of current welding tasks are teleoperated versus autonomous.
• Check whether the robot's size and reach match your specific work envelope, not just a generic 'human-sized' claim.
• Treat 'multipurpose' roadmap claims as separate from the company's current, welding-specific product.

Persona AI in the Broader Humanoid Field

Persona AI sits apart from the general-purpose humanoid race running between companies like Figure, Tesla, and Unitree, covered in our humanoid robot comparison. Rather than compete on breadth of household or warehouse tasks, it is staking a claim on one industrial job with a real customer and a real labor shortage behind it, a more conservative and arguably more testable bet than a general-purpose home robot promise with no paying deployment yet.

Founder backgrounds from NASA's Valkyrie program and Figure AI give the team direct humanoid-hardware experience, a meaningful credibility signal in a field where many entrants are software-first companies retrofitting a humanoid body onto a model built for something else.

The two founder backgrounds also cover different halves of the problem a shipyard deployment actually needs solved. NASA's Valkyrie program built humanoids meant to survive and operate in extreme, unstructured environments with no on-site technician to fix a fault, a design discipline that maps directly onto a hull section full of heat, sparks, and fumes. Figure AI's commercial humanoid work is about manufacturing a humanoid at real production volume rather than as a one-off research platform. A company betting on a narrow industrial niche needs both: hardware that survives the actual job site, and a path to building more than a handful of units once a customer wants a fleet, not a single prototype.