Products

Buy the process that hurts. Add the rest when it pays.

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.

10 process areasPer-process licensingWhole-plant bundle

run_41a9c2 · agent graph · plant BIW-2 · scenario 10 nodes · 1 approval gate
01 · orchestrator ingest.shift ✓ SUCCEEDED 02 · vehicle_line twin.simulate ✓ SUCCEEDED 03 · stamp_form stamp.form ✓ SUCCEEDED 04 · orchestrator approve.human ◆ APPROVAL 05 · weld_join weld.join ✓ SUCCEEDED 06 · frame_geometry frame.geometry ✓ SUCCEEDED 07 · paint_finish paint.finish ✓ SUCCEEDED 08 · powertrain_battery marriage.torque ✓ SUCCEEDED 09 · quality_anomaly inspect.fuse ✓ SUCCEEDED 10 · eol_calibrate eol.release ✓ SUCCEEDED

Illustrative scenario: Plant BIW-2 · EV crossover body shop to end of line · shift 1, one approval gate

The ten

One stack, ten process areas, one record per vehicle

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.

Press shop · stamp and form

Stamp and form

Panel dimension without splits, wrinkles or springback debt.

Body shop · weld and join

Weld and join

Crash-safety joints at rate, without geometry debt.

Framing station · frame and geometry

Frame and geometry

Millimetre bodies at the framing station, not at audit.

Paint shop · paint and finish

Paint and finish

A durable, defect-free finish from the booth, not from repaint.

Pack and marriage · powertrain and battery

Powertrain and battery

Pack and powertrain integrity proven in station, not in the field.

General assembly · trim and fasten

Trim and fasten

Torque, fill and glazing right at the station.

End of line · eol and calibrate

End of line and calibrate

End of line proves the vehicle instead of discovering it.

Plant-wide quality · quality and anomaly

Quality and anomaly

See every defect as it forms, and the station that made it.

Logistics and takt · sequence and flow

Sequence and flow

Takt held against launch cadence, with the buffers to prove it.

Plant twin · vehicle twin

Vehicle-and-line twin

Hit build quality, warranty risk and takt before the shift.

Press shop · agent.stamp_form

Stamp and form

Panel dimension without splits, wrinkles or springback debt.

The problem

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.

What the module does

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.

The agent behind it

Stamp-and-Form: Press force, blank feed, lubrication, die condition, panel dimension.

Key features

  • Press force, stroke, blank feed and lubrication modelled per stroke with closed-loop adjustment inside the qualified window
  • Split, wrinkle and springback risk predicted from material, die and press state
  • Die condition tracked per die set and fed to the maintenance plan
  • Panel-dimension genealogy per blank, handed to Frame-and-Geometry
  • Stamping and springback simulation on the twin before a die or material change

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTDeepStreamTritonOmniverse twinGPU springback FEA surrogate

Why it needs a GPU

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.

MVP scope and first metric

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

  • Press and die-protection interfaces vary by press OEM
  • Springback surrogate fidelity per material grade is to be validated with a design partner
Body shop · agent.weld_join

Weld and join

Crash-safety joints at rate, without geometry debt.

The problem

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.

What the module does

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.

The agent behind it

Weld-and-Join: Spot and laser weld, self-pierce rivet, flow-drill, structural adhesive, robotic joining.

Key features

  • Weld current, force and expulsion traces scored per joint, with electrode-wear and material-lot context
  • Laser seam, rivet, flow-drill and adhesive bead monitored by vision and force
  • Joint-quality and geometry-debt prediction per body, coupled to Frame-and-Geometry
  • Weld schedule and sequence proposals ranked on the twin, approved by a named engineer
  • Robotic joining cells and seam-inspection stations driven inside interlocked limits

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTDeepStreamHoloscanIsaacTritonOmniverse twinGPU weld-thermal FEA surrogate

Why it needs a GPU

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.

MVP scope and first metric

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

  • Servo-gun, laser and robot controller interfaces vary by OEM
  • Qualification evidence required before any weld-schedule write is unattended; the plan keeps it supervised
Framing station · agent.frame_geometry

Frame and geometry

Millimetre bodies at the framing station, not at audit.

The problem

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.

What the module does

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.

The agent behind it

Frame-and-Geometry: Body-in-white dimension, framing accuracy, gap-and-flush.

Key features

  • In-line dimensional data and gap-and-flush vision fused per body
  • Variation predicted from stamping and welding genealogy before the body is out of tolerance
  • Framing and fixture offset corrections inside the qualified envelope
  • Upstream weld-sequence requests to Weld-and-Join, arbitrated by the orchestrator
  • Framing simulation on the twin before a fixture or sequence change

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTHoloscanTritonOmniverse twinGPU point-cloud processing

Why it needs a GPU

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.

MVP scope and first metric

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

  • Metrology and framing-fixture interfaces vary by vendor
  • Correlation between stamping, weld and framing data needs a design partner's full genealogy
Paint shop · agent.paint_finish

Paint and finish

A durable, defect-free finish from the booth, not from repaint.

The problem

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.

What the module does

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.

The agent behind it

Paint-and-Finish: Pretreat and e-coat, sealer, film build, flash and cure, booth environment.

Key features

  • Booth temperature, humidity and airflow controlled to the predicted defect map, inside the qualified window
  • Film build, flash and cure profiles modelled per body and colour
  • Dirt, run and orange-peel defects predicted from booth and chemistry state
  • Paint-defect vision localises each finding to zone, robot and condition
  • Paint-flow simulation on the twin before a recipe or colour change

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTDeepStreamTritonOmniverse twinGPU paint-flow simulation

Why it needs a GPU

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.

MVP scope and first metric

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

  • Booth, oven and paint-robot interfaces vary by paint-system vendor
  • Film-build sensing coverage varies by shop and is confirmed per site
Pack and marriage · agent.powertrain_battery

Powertrain and battery

Pack and powertrain integrity proven in station, not in the field.

The problem

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.

What the module does

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.

The agent behind it

Powertrain-and-Battery: Engine and e-axle build, cell-to-pack joining, coolant and HV integrity, torque-to-yield.

Key features

  • Torque-to-yield traces scored per fastener against the specification
  • Cell-to-pack and cell-to-body joining signals monitored per pack
  • Coolant and HV integrity results fused with the build record
  • Unit holds and routing before a fault leaves the marriage station
  • Proposals to engineers; torque specifications and joint parameters never written by the agent

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTHoloscanTritonNeMo GuardrailsOmniverse twin

Why it needs a GPU

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.

MVP scope and first metric

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-tool, joining and test-stand interfaces vary by vendor
  • Battery-line safety cases are plant specific and set the autonomy level
General assembly · agent.trim_fasten

Trim and fasten

Torque, fill and glazing right at the station.

The problem

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.

What the module does

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.

The agent behind it

Trim-and-Fasten: General-assembly torque, fluid fill, glazing, body-chassis marriage.

Key features

  • Torque traces scored per fastener with tool, operator station and vehicle context
  • Fluid-fill volume and pressure monitored per vehicle and system
  • Glazing bead and placement vision per vehicle
  • Vehicle routing, holds and rework requests before end of line
  • Marriage-station state fused with Powertrain-and-Battery results

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTDeepStreamTritonNeMo Guardrails

Why it needs a GPU

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.

MVP scope and first metric

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

  • Torque-tool and fill-system interfaces vary by vendor
  • Routing writes to the MES depend on the plant's conveyor and buffer design
End of line · agent.eol_calibrate

End of line and calibrate

End of line proves the vehicle instead of discovering it.

The problem

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.

What the module does

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.

The agent behind it

EOL-and-Calibrate: Wheel alignment, ADAS sensor calibration, headlamp aim, leak, brake, roll, functional test.

Key features

  • Alignment, ADAS calibration, headlamp aim, leak, brake, roll and functional results fused per vehicle
  • Failure prediction from upstream weld, geometry, torque and marriage data
  • Retest and rework sequencing against takt
  • Test verdicts and limits never written by the agent
  • First-time-through tracked by station and root cause

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTTritoncuOptRAPIDS

Why it needs a GPU

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.

MVP scope and first metric

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

  • Test-stand and ADAS calibration system interfaces vary by vendor
  • Root-cause coupling needs upstream genealogy from the same plant
Plant-wide quality · agent.quality_anomaly

Quality and anomaly

See every defect as it forms, and the station that made it.

The problem

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.

What the module does

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.

The agent behind it

Quality-and-Anomaly: Weld, geometry, paint, torque, battery and ADAS defects, in real time.

Key features

  • Vision, trace, measurement and test data fused into one build-quality model per vehicle
  • Gap-and-flush and paint-defect vision at the plant edge
  • Defects localised to station, robot, tool and body, handed to the agent that can prevent the next one
  • As-built record and IATF, APQP and PPAP disposition with cited evidence
  • Spec, control-plan and standards Q&A with citations, tenant isolated

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

Jetson OrinTensorRTDeepStreamHoloscanMetropolisTritonNeMoNIMCosmos and Omniverse Replicator (planned)

Why it needs a GPU

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.

MVP scope and first metric

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

  • Synthetic rare-defect generation is planned and needs validation against real distributions
  • Camera coverage per station varies and is confirmed per site
Logistics and takt · agent.sequence_flow

Sequence and flow

Takt held against launch cadence, with the buffers to prove it.

The problem

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.

What the module does

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.

The agent behind it

Sequence-and-Flow: Line sequencing, JIS and JIT feeding, AGV and AMR logistics, takt, line balancing, buffers.

Key features

  • Build sequence, JIS and JIT feeding and buffers optimised with cuOpt against launch cadence
  • AGV and AMR routing coordinated with station demand
  • Line balancing proposals ranked on the twin
  • Sequence and buffer writes to MES and fleet manager inside agreed limits
  • Takt adherence and downtime minutes tracked by cause

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

cuOptRAPIDSTritonOmniverse twinIsaac (AMR autonomy, planned)

Why it needs a GPU

Sequencing, routing and line-balancing optimisation across a plant's variants and fleet, and the telemetry ETL behind it, run on cuOpt and RAPIDS.

MVP scope and first metric

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

  • MES and fleet-manager write interfaces vary by vendor
  • Sequencing constraints per plant are gathered during shadow mode
Plant twin · twin.vehicle_line

Vehicle-and-line twin

Hit build quality, warranty risk and takt before the shift.

The problem

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.

What the module does

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.

The agent behind it

Vehicle-and-Line Twin: As-built vehicle, body and line simulation.

Key features

  • Welding, framing geometry, paint, torque and takt simulated as-built against as-designed
  • Weld-sequence, framing, paint-window and line-balance auto-optimisation handed to the control agents
  • Weld-thermal and geometry FEA, stamping and springback, crash and structural surrogates, paint flow
  • Synthetic rare weld, paint, geometry-drift and misbuild scenarios for validation (planned)
  • Closed learning loop: every real outcome recalibrates the twin

NVIDIA stack

Named in the architecture as the design; items marked planned are not procured

OmniverseOVX (planned)Cosmos and Omniverse Replicator (planned)cuOptDGX or HGX (planned)Modulus-class physics surrogates

Why it needs a GPU

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.

MVP scope and first metric

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

  • Omniverse and OVX infrastructure are planned, not procured
  • Surrogate fidelity per body architecture and material mix is to be validated
Compare

What is in each tier

Areas are licensed per process on the Process tier and bundled on Plant and Enterprise.

What is in each tier
CapabilityProcessPlantEnterprise
Processes covered1All on siteAll in fleet
Agent modules1 of 10 areasAll agentsAll agents + custom models
Vehicle-and-line twinNoIncludedCustom program twins
Plant-edge nodeIncludedPer processFleet managed
Guarded controller writesScoped tagsAll scoped tagsGroup policy
As-built recordIncludedIncludedIncluded + IATF and APQP packets
Autonomy levelsL1 to L2L1 to L3L1 to L4 when qualified
SSO and RBACNoPlannedPlanned, with approval chains
On-prem deploymentNoOptionalIncluded
SupportBusiness hours24/724/7 + named engineer
Indicative price$18,000/mo$140,000/mo$1M to $15M ACV
Run timeline · scenario

The ten areas in one shift

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.

run_41a9c2 Plant BIW-2 · EV crossover body shop to end of line · shift 1 illustrative scenario SUCCEEDED
  1. 01 ingest.shift SUCCEEDED 0.8 s

    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.

  2. 02 twin.simulate SUCCEEDED 1 m 40 s

    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.

  3. 03 stamp.form SUCCEEDED continuous

    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.

  4. 04 approve.human APPROVAL 3 m 12 s

    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.

  5. 05 weld.join SUCCEEDED continuous

    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.

  6. 06 frame.geometry SUCCEEDED continuous

    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.

  7. 07 paint.finish SUCCEEDED continuous

    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.

  8. 08 marriage.torque SUCCEEDED continuous

    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.

  9. 09 inspect.fuse SUCCEEDED continuous

    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.

  10. 10 eol.release SUCCEEDED 2 m 05 s

    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.

Accelerated computing

Physical AI, at the line

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.

Jetson Orin · DeepStream · Holoscan · TensorRT

Edge perception

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.

Triton · NIM

Model serving

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.

DGX or HGX · NeMo

Training

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.

Omniverse · OVX

Plant twin

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.

Cosmos · Omniverse Replicator

Synthetic data

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.

cuOpt

Optimisation

Line sequencing and JIS feeding, AGV routing, takt, line balancing, buffers and retest sequencing.

Isaac · Isaac ROS

Robot cells and AMRs

Robotic weld, paint, assembly and inspection cells and AGV and AMR line-feed autonomy, with skills validated in simulation before a cell runs them.

RAPIDS

GPU ETL

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.

NeMo Guardrails · NVIDIA AI Enterprise

Policy and deployment

Spec, IATF, functional-safety, cyber and human-approval policy enforced before any write; supportable on-prem or plant-edge deployment for regulated OEM plants.

Pricing

Priced per process, per plant, per fleet

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.

Billing period
Process $18,000 per production process or area / month · billed monthly

Weld, paint or assembly control, or defect-inspection AI, for one process. For plants validating ROI on the wedge.

  • One process or area, one agent module
  • Plant-edge inference node included
  • Controller and sensor reads, guarded writes
  • Run log, audit trail, as-built record
  • Business-hours support
Start a pilot
Enterprise Custom land $1M to $15M ACV

Multi-plant OEM and platform fleets standardising on Chassira, with custom weld, paint, geometry and battery models.

  • Multi-plant fleet management
  • Custom weld, paint, geometry and battery models
  • Plant edge fleet with on-prem option
  • Group autonomy policy and approval chains
  • SLAs, SSO and RBAC, dedicated engineer
Contact us
FAQ

Straight answers

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.

Get started

Pick your first process

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.