Technologies
The platforms our engineers can defend in a design review.
Not a logo wall of everything that exists. These are the tools we work in daily and have production systems running on.
Robotics & middleware
ROS 2MoveIt 2GazeboABB RobotStudioFANUC RoboguideKUKA.SimUniversal Robots URScriptStäubli VAL 3Yaskawa MotoSim
Simulation & digital twin
NVIDIA OmniverseIsaac SimIsaac LabOpenUSDPhysXPLCSIM AdvancedEmulate3DPlant SimulationFlexSim
AI & compute
PyTorchTensorRTONNX RuntimeCUDANVIDIA Jetson OrinTriton Inference ServerMLflowHugging FaceLangGraph
Vision
MVTec HALCONOpenCVCognex VisionProKeyenceBaslerGenICamGigE VisionStructured lightLaser profilometry
Controls & fieldbus
Siemens TIA PortalRockwell Studio 5000Beckhoff TwinCATPROFINETEtherNet/IPEtherCATModbus TCPIO-LinkSafety over PROFINET
Mechanical & CAE
SolidWorksCATIA V5Autodesk InventorPTC CreoANSYS MechanicalANSYS FluentEPLAN Electric P8GD&T ISO 1101ASME Y14.5
Data & integration
OPC UAMQTT Sparkplug BIgnitionKepwareInfluxDBTimescaleDBGrafanaAzure IoTAWS IoT SiteWiseISA-95
Engineering practice
PythonC++RustDockerKubernetesGit & GitHub ActionsLinux RT / PREEMPT_RTIEC 62443ISO 13849-1
FIG. — Where each layer sitsSHEET RL-500 · REV B
Architecture
Determinism at the bottom. Context at the top.
The control layer must be deterministic and verifiable — a safety circuit does not get to be probabilistic. The upper layers can be contextual, statistical and adaptive, because their failure mode is a wrong recommendation rather than an unsafe motion.
Most architectural mistakes in industrial AI come from confusing the two: putting a model where an interlock belongs, or a hard-coded rule where a model would have generalised.
Next step
Ready to scope your next automation programme?
Send us a drawing, a cycle-time target or a line layout. Our engineers respond with a technical assessment — not a brochure.