Stamp and form
Panel dimension without splits, wrinkles or springback debt.
Chassira is one platform licensed by process area, not a catalogue of separately named products. The ten areas below map to the plant: press shop, body shop, framing, paint, pack and marriage, general assembly, end of line, plant-wide quality, sequencing and the twin. Start with one on the Process tier and expand to Plant and Enterprise as each proves its number.
Illustrative scenario: Plant BIW-2 · EV crossover body shop to end of line · shift 1, one approval gate
Stamp and form controls the press shop. Weld and join and frame and geometry run the body shop. Paint and finish runs the paint shop. Powertrain and battery and trim and fasten cover marriage and general assembly. End of line proves the vehicle. Quality and anomaly watches every station, sequence and flow holds takt, and the twin rehearses the shift. Each area is a module on the Process tier and all of them are in Plant.
Panel dimension without splits, wrinkles or springback debt.
Crash-safety joints at rate, without geometry debt.
Millimetre bodies at the framing station, not at audit.
A durable, defect-free finish from the booth, not from repaint.
Pack and powertrain integrity proven in station, not in the field.
Torque, fill and glazing right at the station.
End of line proves the vehicle instead of discovering it.
See every defect as it forms, and the station that made it.
Takt held against launch cadence, with the buffers to prove it.
Hit build quality, warranty risk and takt before the shift.
Panel dimension without splits, wrinkles or springback debt.
Body panels are stamped and formed under hard takt from blanks whose material varies by coil. Press force, blank feed, lubrication and die condition drift between checks, and a split, a wrinkle or springback is found at the die or, worse, at framing, where it becomes geometry debt for the whole body. Press shops run on fixed setpoints, periodic checks and the press operator's craft.
The Stamp-and-Form agent reads press force and stroke, blank feed and position, lubricant dose and die condition, predicts splits, wrinkles and springback before the panel leaves the die, and adapts blank feed, lubrication and press settings inside the qualified window. Panel outcomes feed the framing agent and the twin, so a panel's dimension is known before it is welded.
Stamp-and-Form: Press force, blank feed, lubrication, die condition, panel dimension.
Named in the architecture as the design; items marked planned are not procured
Stamping and springback FEA surrogates trained on press telemetry and panel measurement, plus low-latency inference at the press, are GPU workloads. The design target is a prompt under 150 ms for a dimensional anomaly.
Advisory split, wrinkle and springback prediction on one press line from existing press telemetry and panel checks; graduate to bounded blank-feed and lubrication adjustment under approval. First metric: scrap and panel-dimension rework on a named line.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
Crash-safety joints at rate, without geometry debt.
A body-in-white carries hundreds of spot welds, laser seams, self-pierce rivets, flow-drill screws and metres of structural adhesive, every one of them crash-safety relevant, laid down by the plant's largest robot population. Weld parameters and sequences drift with electrode wear, material lots and robot condition, and a bad joint or a geometry-inducing sequence is found at audit, teardown or in the field. Body shops run fixed schedules and sampled destructive tests.
The Weld-and-Join agent reads current, force and expulsion traces from servo guns, laser seam state, rivet and flow-drill force and adhesive bead vision, predicts joint quality and geometry debt per body, and proposes weld schedules, joining sequences and robotic joining parameters. Because weld parameters are crash-safety critical, every schedule change waits for a named body-shop engineer until the joint family is qualified. The Robot-and-Inspect agent drives the cells and seam-inspection stations.
Weld-and-Join: Spot and laser weld, self-pierce rivet, flow-drill, structural adhesive, robotic joining.
Named in the architecture as the design; items marked planned are not procured
Multi-stream seam vision, high-rate weld-trace fusion and weld-thermal and geometry FEA surrogates are GPU workloads. The design target is a weld or gap alert under 50 ms at the cell.
The leading wedge in the plan: AI weld-defect and gap-and-flush vision in one body shop, shadow mode first, then schedule proposals approved by the launch engineer. First metric: weld rework per body, or expulsion and geometry-debt events on a named line.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
Millimetre bodies at the framing station, not at audit.
Body-in-white dimension and gap-and-flush are decided at framing, where panels, subassemblies and fixtures meet. Variation accumulates from stamping and welding and is found by in-line measurement, CMM audit or the customer's eye. A body out of tolerance costs rework, downtime and warranty, and a framing drift costs a shift of bodies before anyone acts.
The Frame-and-Geometry agent reads in-line dimensional points, framing-station offsets and gap-and-flush vision, predicts where variation is heading, and proposes framing and fixture corrections and upstream sequence changes so bodies hit millimetre targets at the station. Corrections inside the qualified envelope run under the plant's autonomy level; anything larger goes to the launch engineer.
Frame-and-Geometry: Body-in-white dimension, framing accuracy, gap-and-flush.
Named in the architecture as the design; items marked planned are not procured
Point-cloud and photogrammetry processing from in-line metrology and framing-geometry FEA surrogates are GPU workloads. The design target is a dimensional-anomaly prompt under 150 ms.
Advisory geometry-drift prediction on one framing station from existing in-line measurement; graduate to bounded offset corrections under approval. First metric: bodies out of tolerance per shift, or gap-and-flush rework on a named line.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
A durable, defect-free finish from the booth, not from repaint.
Paint shops coat millions of square metres a year through pretreat, e-coat, sealer, primer, base and clear, flash and cure, inside booths whose temperature, humidity and airflow drive dirt, runs, orange peel and film-build faults. Defects are found at inspection and fixed by repaint, scrap and corrosion warranty. Booth environments and oven profiles run on fixed recipes and the paint engineer's craft.
The Paint-and-Finish agent reads booth environment, film build, flash and oven profiles and pretreat and e-coat chemistry, predicts the defects a body will carry before it reaches inspection, and adapts booth environment, flash and cure setpoints inside the qualified window. The Quality-and-Anomaly agent's paint-defect vision closes the loop by localising each defect to zone and condition.
Paint-and-Finish: Pretreat and e-coat, sealer, film build, flash and cure, booth environment.
Named in the architecture as the design; items marked planned are not procured
Paint-defect vision across many cameras per booth and paint-flow and film-build simulation are GPU workloads. The design target is a paint-anomaly prompt under 150 ms.
A plan wedge: paint-defect detection and film-build control in one paint shop, shadow mode first, then booth-environment adjustments under approval. First metric: repaint rate, or paint defects per body by type on a named shop.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
Pack and powertrain integrity proven in station, not in the field.
Engine and e-axle build, cell-to-pack and cell-to-body joining, coolant and high-voltage integrity and torque-to-yield fastening are the most safety-critical and least visible steps in the plant, and the EV transition is standing up new battery lines faster than engineers can staff them. A joint fault, a torque error or a coolant leak that escapes marriage becomes a warranty claim or a recall.
The Powertrain-and-Battery agent reads torque-to-yield traces, cell-to-pack joining signals, coolant and high-voltage integrity tests and marriage-station telemetry, predicts joint and pack-integrity faults per unit, and holds or routes units before they leave the station. It never changes a torque specification or a joint parameter; it proposes, and a named engineer decides.
Powertrain-and-Battery: Engine and e-axle build, cell-to-pack joining, coolant and HV integrity, torque-to-yield.
Named in the architecture as the design; items marked planned are not procured
Fusing high-rate torque, joining and integrity signals with vision across a pack line, and training the fault models behind it, are GPU workloads. The design target is a prompt under 150 ms.
A plan wedge: torque and battery-joint conformance on one pack or marriage line, shadow mode first, then holds and routing under approval. First metric: faults caught before end of line, or marriage rework on a named line.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
Torque, fill and glazing right at the station.
General assembly runs thousands of torque-controlled fasteners, fluid fills, glazing operations and the body-chassis marriage per shift. A mis-torque, an under-fill or a glazing fault escapes to end of line or to the customer, and the andon reaction stops the line at tens of thousands of dollars a minute. Traceability systems record the fault; nobody prevents the next one.
The Trim-and-Fasten agent reads torque traces per fastener, fluid-fill volumes and pressures, glazing vision and marriage-station state, predicts mis-torque, under-fill and glazing faults per vehicle, and routes or holds vehicles before the fault reaches end of line. Fill setpoints inside the qualified window may move at the plant's autonomy level; torque specifications stay with the process engineer.
Trim-and-Fasten: General-assembly torque, fluid fill, glazing, body-chassis marriage.
Named in the architecture as the design; items marked planned are not procured
Per-fastener trace scoring across a general-assembly line and glazing vision at rate are GPU workloads. The design target is a torque or fill prompt under 150 ms.
Advisory torque and fill fault prediction on one general-assembly zone from existing traceability data; graduate to routing and holds under approval. First metric: torque and fill rework per vehicle, or faults escaping to end of line.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
End of line proves the vehicle instead of discovering it.
Wheel alignment, ADAS sensor calibration, headlamp aim, leak, brake, roll and functional tests are the last gate before the customer, and the ADAS content per vehicle keeps growing. Failures are discovered at the test stand, retested and reworked under takt, and a mis-calibrated sensor that escapes is a safety recall.
The EOL-and-Calibrate agent reads every end-of-line result per vehicle, predicts calibration and test failures from the upstream build record, and sequences retests and rework so the stand proves the vehicle. It never passes a vehicle a test has failed and never changes a test limit.
EOL-and-Calibrate: Wheel alignment, ADAS sensor calibration, headlamp aim, leak, brake, roll, functional test.
Named in the architecture as the design; items marked planned are not procured
Fusing end-of-line results with the full upstream build record per vehicle and training failure-prediction models across programs are GPU workloads. Retest sequencing runs on cuOpt.
Advisory end-of-line failure prediction on one line from existing test-stand data; graduate to retest sequencing under approval. First metric: first-time-through rate, or ADAS and alignment retests per shift.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
See every defect as it forms, and the station that made it.
Weld, geometry, paint, torque, battery and ADAS defects are found by sampled inspection, audit and andon reaction, each in its own system. Plant managers run blind on real-time build quality across body, paint, powertrain and final, and the launch engineer's knowledge of why a defect appears lives in a few people.
The Quality-and-Anomaly agent fuses weld traces, dimensional data, gap-and-flush and paint-defect vision, torque traces and battery and ADAS results into one model of build quality per vehicle, detects and predicts defects as they form, and localises each to station, robot, tool and body. The Quality-and-Conformance agent writes the as-built record and dispositions against IATF 16949, APQP and PPAP with cited evidence, and the Vehicle-Knowledge agent answers spec and standards questions with citations.
Quality-and-Anomaly: Weld, geometry, paint, torque, battery and ADAS defects, in real time.
Named in the architecture as the design; items marked planned are not procured
Multi-camera gap-and-flush and paint-defect vision, high-rate trace fusion and synthetic rare-defect generation are GPU workloads. The design targets are a weld or gap alert under 50 ms and a paint or dimensional prompt under 150 ms.
The inspection half of the leading wedge: gap-and-flush or paint-defect vision on one line in shadow mode, compared against audit. First metric: inspection escapes, or defects caught in station rather than at audit.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
Takt held against launch cadence, with the buffers to prove it.
Line sequencing, just-in-sequence and just-in-time feeding, AGV and AMR logistics, line balance and buffers are tuned by hand against a launch cadence that keeps changing. A sequence violation or a starved station costs downtime minutes at tens of thousands of dollars each, and the EV and software-defined-vehicle transition multiplies the variants to sequence.
The Sequence-and-Flow agent reads the build sequence, feeding status, AGV and AMR fleet state, station cycle times and buffers, prices every sequencing and buffer move in takt, downtime and rework, and optimises sequence, feeding and routing with cuOpt. Writes go to the MES and fleet manager inside the limits the plant sets.
Sequence-and-Flow: Line sequencing, JIS and JIT feeding, AGV and AMR logistics, takt, line balancing, buffers.
Named in the architecture as the design; items marked planned are not procured
Sequencing, routing and line-balancing optimisation across a plant's variants and fleet, and the telemetry ETL behind it, run on cuOpt and RAPIDS.
Advisory sequencing and buffer recommendations on one line from MES and fleet data; graduate to bounded sequence writes under approval. First metric: takt adherence and downtime minutes, or sequence violations per shift.
Tier. Process tier, $18,000 per production process or area per month, expanding to Plant and Enterprise.
Open questions
Hit build quality, warranty risk and takt before the shift.
Whether a body hits geometry, a finish is defect free, a pack is torqued to spec and a line holds takt depends on a coupled chain from stamping to end of line that is today tuned by trial builds, offline simulation and launch craft. Launch ramps burn bodies and weeks, and plant managers cannot see the quality, warranty and takt trade-offs before committing the shift.
The Vehicle-and-Line Twin simulates welding, framing geometry, paint, torque and takt on an Omniverse plant twin, compares as-built against as-designed, and auto-optimises the weld sequences, framing offsets, paint windows and line balance the other agents execute. Weld-thermal and geometry FEA, stamping and springback, crash and structural surrogates and paint-flow simulation run on GPUs so the shift is rehearsed before it starts. The pattern is production proven at BMW and Mercedes-Benz, as the architecture notes; Chassira's twin is a design.
Vehicle-and-Line Twin: As-built vehicle, body and line simulation.
Named in the architecture as the design; items marked planned are not procured
Plant-twin simulation, FEA surrogates and photogrammetry are the compute-heavy workloads that make the twin GPU essential. The design target is 50,000 to 250,000 synthetic images or measurement variants per vehicle program before live deployment.
Sold inside the Plant tier. First use: as-built simulation and pre-shift prediction for one body shop, ranking weld-sequence and framing candidates for the launch engineer. First metric: launch-ramp quality and takt against the program plan.
Tier. Sold inside the Plant tier, $140,000 per plant per month, and in Enterprise.
Open questions
Areas are licensed per process on the Process tier and bundled on Plant and Enterprise.
| Capability | Process | Plant | Enterprise |
|---|---|---|---|
| Processes covered | 1 | All on site | All in fleet |
| Agent modules | 1 of 10 areas | All agents | All agents + custom models |
| Vehicle-and-line twin | No | Included | Custom program twins |
| Plant-edge node | Included | Per process | Fleet managed |
| Guarded controller writes | Scoped tags | All scoped tags | Group policy |
| As-built record | Included | Included | Included + IATF and APQP packets |
| Autonomy levels | L1 to L2 | L1 to L3 | L1 to L4 when qualified |
| SSO and RBAC | No | Planned | Planned, with approval chains |
| On-prem deployment | No | Optional | Included |
| Support | Business hours | 24/7 | 24/7 + named engineer |
| Indicative price | $18,000/mo | $140,000/mo | $1M to $15M ACV |
Every Chassira run is an ordered, inspectable sequence. This is run_41a9c2 on Plant BIW-2 · EV crossover body shop to end of line · shift 1. Scenario completes in 10 steps with one approval gate. Weld schedule change on robot W-14 approved by the body-shop launch engineer, framing offset held inside the envelope, booth humidity trimmed before the predicted dirt defects appeared, one pack held at marriage on a torque trace, and the shift released with full as-built records.
Plant Orchestrator · Pulled the shift build sequence, program control plan and qualified envelopes from MES: 412 bodies, four trim variants, weld schedule set W-rev 7, framing targets per station, paint colour blocks, torque specifications. Locked every safety-critical parameter as read-only.
Vehicle-and-Line Twin · Simulated 18 candidate weld-sequence, framing-offset and booth-window combinations on the plant twin, ranked on predicted body geometry, expulsion risk at robot W-14, paint defect map and takt against launch cadence.
Stamp-and-Form · Blank feed on press line P-3 adjusted by 0.4 mm and lubricant dose raised 3 percent for the new coil lot as springback risk on the rear door inner climbed. Panel dimension held inside the window; genealogy handed to framing.
Plant Orchestrator · Candidate #6 lowers weld current 2 percent and adds 20 ms hold on robot W-14 to cut predicted expulsion. Weld schedules are crash-safety critical and stay with a human, so the change was held for the body-shop launch engineer and approved at 06:14:52.
Weld-and-Join · Approved schedule 7b applied on W-14. Current and force traces confirmed inside the window on the first 40 bodies; predicted expulsion risk fell from 0.31 to 0.06 and no geometry debt was introduced at the adjacent seams.
Frame-and-Geometry · Framing station FS-1 offset corrected by 0.3 mm on the left B-pillar datum inside the qualified envelope as in-line measurement trended toward the tolerance edge. Gap-and-flush vision confirmed the correction on the next 25 bodies.
Paint-and-Finish · Booth B-2 humidity trimmed 2 points and airflow rebalanced for the dark-colour block as the defect model predicted dirt on horizontal panels. Film build held inside the window on the first colour block.
Powertrain-and-Battery · Pack PK-1182 held at marriage after a torque-to-yield trace on fastener 7 fell outside the specification curve. Routed to rework with the trace attached; torque specification untouched, as always.
Quality-and-Anomaly · Gap-and-flush and paint-defect cameras, weld and torque traces and test results fused per body. Two paint dirt defects localised to booth B-2 zone 3 before the humidity trim took effect; both routed to spot repair, none escaped to final.
EOL-and-Calibrate · End of line proved alignment, ADAS calibration, leak, brake, roll and functional tests on every vehicle; two ADAS retests sequenced inside takt. Shift 1 released with as-built records, approvals and dispositions written for the quality lead.
A spot weld lasts a fraction of a second and a body passes a station every minute. Perception has to be local, deterministic and fast, so Chassira runs GPU inference at the plant edge and keeps training, simulation and optimisation in the cloud or on-prem.
8 to 32 cameras per process area at 15 to 60 FPS, plus torque, dimensional, thermal, robot and paint-instrument streams fused at the cell. Weld and gap alerts under 50 ms and paint or dimensional prompts under 150 ms are the design targets. Jetson Thor is planned for higher density.
Weld-defect, geometry-risk, paint-risk, torque and battery, and takt-risk models served at the plant edge with versioned rollback, routed by latency, tenant and cost, with canary rollout by line.
Weld, paint and gap vision, dimensional models, NeMo IATF and spec reasoning and GPU-accelerated stamping, springback and weld-geometry FEA surrogates. 150 to 600 GPU-hours per plant during onboarding, then monthly retraining from engineer corrections. Dedicated capacity is planned.
Body framing, weld sequence, robot reach, paint flow and line balance simulated before the shift. The pattern is production proven at BMW and Mercedes-Benz, per the architecture; Chassira's twin and OVX capacity are planned.
Rare weld defects, paint failures, geometry-drift modes, occluded joints and misbuild cases generated and validated against real distributions before promotion. 50,000 to 250,000 variants per program is the design target. Planned.
Line sequencing and JIS feeding, AGV routing, takt, line balancing, buffers and retest sequencing.
Robotic weld, paint, assembly and inspection cells and AGV and AMR line-feed autonomy, with skills validated in simulation before a cell runs them.
Telemetry ETL across weld and press parameters, measurements, paint readings, torque traces, battery and ADAS results, non-conformance and takt history for retraining and benchmarks.
Spec, IATF, functional-safety, cyber and human-approval policy enforced before any write; supportable on-prem or plant-edge deployment for regulated OEM plants.
Land on one process with a measured baseline. Expand by process, plant and autonomy level. Outcome components are available on rework, downtime, warranty and first-time-through. Annual prepay carries a 15 to 20% discount.
Weld, paint or assembly control, or defect-inspection AI, for one process. For plants validating ROI on the wedge.
The whole plant: stamp and form, weld and join, frame and geometry, paint and finish, powertrain and battery, trim and fasten, end of line and calibrate, quality and anomaly, sequence and flow, and the twin.
Multi-plant OEM and platform fleets standardising on Chassira, with custom weld, paint, geometry and battery models.
The questions plant managers, launch engineers and quality leads ask in the first meeting.
Yes, but only within an explicit tag allow-list with per-tag rate and magnitude limits, inside the qualified envelope, and only at the autonomy level your plant has set. Design partners start in shadow mode, where Chassira writes nothing. Weld-schedule changes wait for a named body-shop engineer, and torque specifications, battery-joint parameters and end-of-line test limits and verdicts are read-only to every agent, always.
Control returns to the press, robot, booth or MES controller's last known-good state. Chassira is a supervisory layer on top of the safety-rated PLCs and interlocks you already run, never a replacement for them, so an outage degrades the line to its current way of running, not to a stop.
Advisory recommendations start from the plant twin and the control plan in the first weeks. Weld, geometry, paint, torque and end-of-line prediction improve as line history and labelled outcomes accumulate per program. The compute plan sizes onboarding at 150 to 600 GPU-hours per plant and 50,000 to 250,000 synthetic variants per program, all design targets.
Only if you choose cloud training. Designs, control plans, specs, process models and quality records are isolated per tenant and never used to train another OEM's models. On-prem and plant-edge deployment are available for regulated plants, and federated cross-plant learning is planned and waits on data-rights agreements.
You are, the same as with any control strategy, which is why every write is policy-checked, bounded, logged and reversible, and why anything crash-safety or IATF relevant waits for a named approver. The audit log records the request, the evidence, the envelope applied and the human decision.
A paid pilot on one process or area at the Process tier, scoped to a single wedge with a success metric written into the order form: weld or geometry rework, downtime minutes, paint defect or repaint rate, warranty or PPM, or launch quality. Shadow mode first, then assist mode with your engineer approving each move, then bounded autonomy on low-risk loops once ROI and safety gates are met.
Start with one process and one measurable baseline. A paid pilot on body-shop weld and gap-and-flush vision, paint-defect detection and film-build control, or torque and battery-joint conformance shows the number before you commit further.
↑↓ navigate↵ openesc close