When humanoids move from demos to production lines, the problem changes
Humanoid news tends to focus on whether a robot can walk naturally, manipulate objects or understand instructions. Those questions matter at the prototype stage. But once the same machine has to be built hundreds or thousands of times, a different engineering problem appears: making one robot work is not the same as producing every robot at consistent quality.
Figure said in April 2026 that it had produced more than 350 Figure 03 robots and increased throughput from one robot per day to one per hour in under 120 days. The company described the biggest fundamental challenge in that ramp not as the AI model, but as improving yield rate and cycle times.
One robot per hour still requires more than 50 process inspections and 80 final tests
Figure's BotQ runs more than 150 networked workstations and more than 50 in-process inspection points. Each completed robot is subjected to more than 80 functional verification tests, including multi-limb stress tests and burn-in sessions with thousands of cycles of squats, shoulder presses and jogging.
Figure reported end-of-line first-pass yield above 80% and improving weekly, a 99.3% first-pass yield on the battery line with more than 500 packs shipped, and more than 9,000 actuators produced across over 10 SKUs. These are self-reported company metrics, not independently audited industry statistics, but they clearly show where engineering effort is being allocated.
Quality now stretches from suppliers to field failures
Figure says it qualifies hundreds of suppliers against incoming inspection criteria, checks quality throughout assembly, conducts end-of-line testing and feeds field failures back into engineering for hardware revisions.
That resembles the closed-loop quality systems long used in automotive, electronics and industrial equipment: Supplier Quality → Incoming Quality → In-process QC → End-of-Line Test → Reliability → Field Failure Analysis. An AI-powered robot does not escape traditional manufacturing quality. It adds software and learning systems on top of it.
The job board shows quality and reliability becoming organizations of their own
Figure's current BotQ openings include Manufacturing Engineers, Manufacturing Test, Quality Engineering Technicians, Senior Quality Engineers and Supplier Quality Engineers. Its Senior Quality Engineer role explicitly targets manufacturing yield and asks for SPC, 8D, 5-Why, Fishbone analysis, vendor audits and corrective actions.
Supplier Quality works across injection molding, die casting, stamping, wire harnesses and textiles, while Reliability Technicians perform component-, module- and final-product testing and failure analysis. Figure had already said when introducing BotQ that it had stood up dedicated safety and reliability teams as it moved toward production.
Tesla Optimus is building a factory workforce next to its AI workforce
Tesla's current Optimus listings include Sr. Quality Engineer, Quality Supervisor, Supplier Quality Engineer, Manufacturing Test Engineer, General Assembly & Testing, NPI, Repair & Validation and other production roles.
The Sr. Quality Engineer - Optimus posting specifically focuses on the launch phase of humanoid production, supplier readiness and continual improvement of systems that control product quality. Uncertainty remains around Tesla's actual Optimus production schedule, but the public hiring evidence is enough to show that the workforce is not being built around AI model development alone.
This is where automotive, semiconductor and electronics talent becomes adjacent to robotics
SPC, 8D, 5-Why, FAI, DFM/DFA, metrology and end-of-line testing were not invented by robotics. They are mature manufacturing capabilities from automotive, semiconductor, consumer electronics and industrial equipment.
That makes Quality Engineers, Supplier Quality Engineers, Manufacturing Engineers, Test/Validation Engineers and Reliability Engineers plausible adjacent talent for Physical AI. Adjacency is not equivalence, however. Humanoids combine high-degree-of-freedom actuators, batteries, sensors, embedded electronics, safety and software, so evidence of working with complex electromechanical systems still matters.
Japan and Korea should look beyond the shortage of 'robot AI' talent
Japan and Korea already have deep talent pools in automotive, batteries, precision components, semiconductors, factory automation and industrial robotics. Looking at Physical AI only through the number of AI researchers understates this industrial base. As production scales, supplier qualification, process capability, metrology, test automation and field reliability all become critical.
Figure is currently recruiting a Supplier Quality Engineer (Asia), which is relevant as a supply-chain signal. The posting does not specify a country, so it should not be interpreted as direct hiring in Korea or Japan. The narrower conclusion is that humanoid supply-chain quality is already extending beyond U.S. headquarters into Asian supplier networks.
The next competition is between one smart robot and a thousand robots of the same quality
AI models, data, autonomy and manipulation remain central. But once production begins, competition no longer stays inside the model. A prototype that works once and a fleet of thousands that performs consistently and safely are different engineering achievements.
Figure's production metrics and the current Figure and Tesla job boards show that this second problem is already creating teams and jobs. As Physical AI moves from research into industry, some of the most valuable people may be those who find failures, remove root causes and make the process repeatable.
BANSEOG VIEW
The missing part of the Physical AI talent map may sit between prototype and production, not between AI research and robot development
As humanoids move toward volume production, talent competition expands from AI and robotics researchers into supplier quality, manufacturing test, reliability, failure analysis, NPI and field service. Figure's production ramp and current Figure and Tesla hiring provide public evidence that this transition has begun.
Banseog Search views the opportunity through capability transfer rather than job titles. SPC, 8D, FAI, DFM/DFA, metrology, end-of-line testing and reliability expertise built in automotive, semiconductor, electronics and industrial equipment may become increasingly relevant to real humanoid hardware.
SOURCES
Primary sources and references
- Figure · Ramping Figure 03 Production (Apr. 29, 2026)
Figure self-reported 350+ robots, 1/day→1/hour, 50+ inspections, >80% EOL FPY, 99.3% battery FPY and 80+ verification tests.
- Figure · Senior Quality Engineer, BotQ
Current role focused on yield improvement, SPC, 8D, 5-Why, Fishbone, supplier audits and corrective action.
- Figure · Supplier Quality Engineer, BotQ
Current role covering supplier qualification, FAI, DFM and root-cause analysis.
- Figure · Supplier Quality Engineer (Asia)
Current Asia supplier-quality role. Not interpreted as direct hiring in any specific country.
- Figure · Reliability Technician
Current role performing reliability testing and failure analysis from prototype through production.
- Tesla · Sr. Quality Engineer, Optimus
Current role focused on launch-phase Optimus production quality and supplier readiness.
- Tesla · Optimus careers
Current listings include quality, supplier quality, manufacturing test, general assembly and related production roles.
Figure's production metrics are company-reported and not independently audited industry statistics. Figure and Tesla job postings describe current demand at specific companies and do not establish the size of the humanoid labor market or a long-term wage premium. No private Banseog candidate, client or Search evidence is used.