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
The problem
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.
The solution
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.
Capabilities
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
In the 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.
Autonomy
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.
Outcomes
What Roboq moves
0First-pass yieldin the loop
0Fewer false callsvs. rule-based
0Less scrap & rework
0Edge inferenceper tile
Built on NVIDIA
Technology overview
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.
Questions
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.
Every board, perfectly built.
See Roboq run on your boards
Book a working session and watch Roboq perceive, predict, and close the loop against your line data.