Vision-guided handling
Adaptive board loading, magazine, and tray handling.
Physical Actuation · Circuex agent 06
Roboq
Physical-AI robotics for the electronics factory. Board handling, depaneling, and box-build that adapt to any product, trained in Isaac Sim, no per-SKU re-fixturing.
Act & handleGraduated autonomyMES write-back
LINE VIEW · PHYSICAL ACTUATION
99.4%FPY
20–50msEdge inference
LiveClosed loop
AUTOMOTIVEMEDICALAEROSPACEINDUSTRIALAI HARDWAREEMS / CM
Why this stage bleeds
Between and after SMT, boards are loaded, flipped, depaneled, and assembled into enclosures, much of it manual or on rigid fixtures re-tooled per product. High-mix makes fixed automation uneconomic and risks damage to dense, fragile boards.
How Roboq closes it
Roboq brings adaptive, vision-guided robotics to loading, magazine handling, depaneling, flipping, and box-build, trained and validated in Isaac Sim so it generalizes across boards and enclosures without re-fixturing.
What Roboq does
Depaneling & flip
Singulation and buffering without hard fixtures.
Box-build assembly
Connector/cable insertion, fastening, enclosure close-up.
Sim-first skills
Training and validation in Isaac Sim on product CAD.
ESD-aware manipulation
Force and vision feedback that avoids board damage.
Loop coordination
Handling respects quality state and traceability.
LINE VIEW · PHYSICAL ACTUATION
99.4%FPY
20–50msEdge inference
LiveClosed loop
Roboq never works blind
It shares perception, twin validation, and orchestration with the rest of Circuex, so every action is grounded and every result feeds the next board.
- Act & handle: its role in the perceive–simulate–decide–act loop.
- Twin-validated: moves proven in Twinex before production.
- Cited & logged: assurance-grade audit trail on every change.
Graduated, never reckless
You choose how much control Roboq takes, and it earns the next level by proving itself.
01
Shadow
Roboq perceives and predicts with zero write-back. You validate its calls against reality.
02
Recommend
It surfaces grounded recommendations with the evidence behind each one.
03
Approve
Operators approve actions one at a time under policy, building trust and data.
04
Closed-loop
Roboq tunes autonomously within guardrails, escalating only exceptions.
What Roboq moves
99.4%First-pass yieldin the loop
72%Fewer false callsvs. rule-based
40%Less scrap & rework
20msEdge inferenceper tile
The physical-AI stack behind Roboq
Trained on DGX/HGX, served at the edge on Jetson and IGX, validated in Omniverse and Isaac Sim.
JetsonEdge inference
TensorRTLow-latency serving
OmniverseTwin & synthesis
PHYSICAL ACTUATION · EDGE
20–50msPer tile
8–16Cameras
LiveOn line
Once Roboq went closed-loop, the stage stopped being the bottleneck. It just holds yield, and tells us why when it doesn't.
Roboq FAQ
Does Roboq require the other agents?
No. It's a standalone product with its own ROI, but it compounds when paired with the rest of the Circuex loop.
How does it connect to our equipment?
Read-only connectors first, then graduated write-back to the relevant machines and your MES/traceability system.
Is it safe for high-reliability production?
Yes, graduated autonomy, approval policies, and an assurance-grade audit trail are built in for automotive, medical, and aerospace.
See Roboq run on your boards
Book a working session and watch Roboq perceive, predict, and close the loop against your line data.