Mission
To make robotics perception deployable in a day — by shipping a unit that turns the cameras a robotics team already chose into dense 3D, tracked objects, and scene understanding, without that team having to stand up an EdgeAI program first.
Sovereign robotics vision
Invotet Technology Inc.
Invotet builds the RVPU — a Robotics Vision Processing Unit. It is deliberately not a generic AI accelerator. It is a curated set of robotics perception functions in a hardware package: a ground-robotics team wires its cameras in and gets dense stereo 3D, object detection and segmentation, 3D mapping and temporal tracking, and cross-modality matches back. Perception as a unit — not "compile your own models onto an accelerator and build your own pipeline."
Vision
To become the sovereign, resilient perception layer for autonomous machines — a future where a robot can see and understand its environment on-platform, in contested and comms-denied conditions, without depending on a cloud round-trip or on another nation’s silicon supply.
Story
What Invotet does, briefly.
Invotet builds the RVPU — a Robotics Vision Processing Unit. It is deliberately not a generic AI accelerator. It is a curated set of robotics perception functions in a hardware package: a ground-robotics team wires its cameras in and gets dense stereo 3D, object detection and segmentation, 3D mapping and temporal tracking, and cross-modality matches back. Perception as a unit — not "compile your own models onto an accelerator and build your own pipeline."
The problem we solve is design freedom. An engineer building a robot works under tight SWaP, latency, cost, and supply-chain constraints, but off-the-shelf 3D platforms force a vendor camera kit that fixes baseline, placement, and field of view. The alternative — designing your own stereo rig or a multi-camera 360° setup — is custom engineering work most teams cannot afford to keep reinventing. The RVPU takes the cameras and the geometry you chose and does the perception.
Perception runs on-platform, on FPGA-accelerated engines, and keeps working when the link drops. Outputs land in the shapes robotics stacks consume. A hardware root of trust and per-module attestation keep the device tamper-evident, and an 8,192-entry cycle-accurate trace buffer makes every run reviewable. Sovereign and OEM programs can license the SDK and IP to keep a perception stack fully under their own control.
Headquartered in Decatur, Georgia.
Operating principles
Four pillars under the RVPU.
Pillar 01
Vision at the Edge
Perception runs on the platform — minimizing latency, eliminating downlink dependency, and keeping data inside the device.
Pillar 02
No EdgeAI Team Required
Configured perception functions instead of a model-compile path. Robotics teams deploy working vision by configuring a unit, not by staffing an AI and FPGA program.
Pillar 03
Design Freedom
Your cameras, your baseline, your layout. The unit adapts to the rig your platform needs rather than forcing a locked vendor kit that fixes the decisions that matter most.
Pillar 04
Sovereign & Resilient
Sovereign supply and defense-grade posture, deployable under any country’s sovereignty requirements. Perception keeps working in contested, comms-denied environments where a cloud round-trip is not an option.
Headquarters
Decatur, GA
2052 N Decatur RoadDecatur, GA 30033
USA