Push-button deployment.
Set up your model and target, then launch one automated workflow from compilation through to inference on supported hardware.
INTRODUCING VERTIS
Push-button
deployment.
From your AI model to a running FPGA.
Vertis automates compilation, verification, programming and inference in one push-button workflow for supported models and targets.
FOR AI & EMBEDDED ENGINEERING TEAMS
VERTISMODEL → HARDWARE
01 / INPUTyour_model.onnx
ONNX02 / COMPILATION & VERIFICATION
Model-aware.Target-aware.
FPGASUPPORTED TARGET
03 / PROGRAM & RUN
+PUSH-BUTTON MODEL-TO-HARDWARE DEPLOYMENT+
VENDOR AGNOSTIC BY DESIGN.↓
TRUSTED BY
01 / MEET VERTIS
A trained model is
only the beginning.
Between a model and working hardware sits a complex chain of engineering decisions. Vertis turns the supported deployment path into a push-button workflow.
Import your ONNX model, define the target and constraints, then start deployment. Vertis handles the supported workflow through compilation, verification, FPGA programming and inference.
Vertis is built by AI and FPGA engineers who have experienced the challenges of deploying AI on FPGAs firsthand. We built it to simplify the workflow we needed ourselves.
Vendor independent at the core.
A target-neutral compiler with separate vendor backends, designed to connect with different FPGA toolchains.
Evidence along the way.
Keep model identity, build settings, generated artefacts and verification outcomes together.
02 / THE WORKFLOW
You define it.
Vertis deploys it.
One push-button workflow. Four connected stages.
Explore what Vertis automates for supported deployments.
START WITH YOUR MODELThe model is yours.
The model is yours.
The path starts here.
Bring an ONNX inference model. Vertis inspects its graph, inputs and operators, then connects it to the intended hardware target and deployment settings.
MODEL + TARGET + CONSTRAINTS
⌘your_model.onnxINFERENCE MODEL
SourceONNX graph
TargetExplicitly selected
ScopeSupported operators & shapes
Working on real hardware. The supported automated workflow has been demonstrated with multiple AI models.
See current scope03 / VENDOR-AGNOSTIC ARCHITECTURE
One workflow.
Different vendors.
Vertis is vendor agnostic by design. A common compiler connects to separate vendor backends, keeping your deployment workflow consistent across FPGA toolchains.
Each backend handles its vendor’s synthesis, implementation and programming tools. Vertis brings them into one push-button deployment path.
Apply with your targetYOUR ONNX MODELYOUR TARGET & CONSTRAINTS
VENDOR-AGNOSTIC COMPILERVendor ABACKENDToolchainFPGA
Vendor BBACKENDToolchainFPGA
Vendor CBACKENDToolchainFPGA
ENGINEERING YOU CAN INSPECT
Know what
went into it.
And what came out.
A deployment is more than a generated file. Vertis records the context behind each build, so engineers can inspect the artefacts and the checks that produced them.
01Source model & identity↗
02Target & build configuration↗
03Generated deployment artefacts↗
04Separate verification outcomes↗
04 / BUILT FOR THE EDGE
Intelligence, closer
to where it matters.
For teams developing AI within real power, compute and connectivity constraints.
01 / SPACE SYSTEMS
Process closer to the sensor.
Explore onboard inference for applications where downlink, power and compute are limited.
02 / INDUSTRIAL EDGE
Bring AI into the system.
Connect the model to a hardware deployment path shaped around the application’s constraints.
03 / AUTONOMOUS SYSTEMS
Keep inference local.
Develop onboard intelligence for systems operating beyond dependable remote connectivity.
05 / THE ENGINEERING BEHIND VERTIS
AI minds.
Hardware instincts.
Setanta brings together spacecraft AI, FPGA systems and embedded engineering. Alongside Vertis, our work spans onboard intelligence and modular computing for the mission.
Explore the applications we work on and the hardware we build around them.
Built in Dublin. Designed for the edge.
CONCEPT / ONBOARD INFERENCE
AI & ONBOARD INTELLIGENCE
Make more of
what you sense.
From Earth observation to navigation and sensor alerts. Shape AI workloads around the decisions your system needs to make.
Explore AI
DANU / MODULAR ONBOARD COMPUTE
ONBOARD COMPUTERS & SPACE TILES
Build around
your mission.
A modular hardware range for spacecraft computing, with FPGA, processing and memory options for mission-specific capability.
Explore hardwareA CLEAR VIEW OF THE PRODUCT
Start with
the right scope.
Vertis is under active development.
Let’s define an evaluation around your model and hardware.
What has been demonstrated?
The supported automated workflow - model import, compilation, FPGA programming and inference - has been demonstrated using multiple off-the-shelf AI models on AMD FPGA hardware.
Does Vertis support every FPGA?
Vertis has a vendor-independent integration architecture. This does not mean every device, model or toolchain version is validated. Additional backends and target validation are being developed; we assess your intended configuration with you.
Can I bring my own model?
Yes: ONNX is the canonical model format. Support depends on the graph, operators, shapes, numerical representation and target. We can discuss a bounded evaluation of your inference model and the criteria that matter to your team.
THE VERTIS CLOSED ALPHA
Build with
Vertis.
Help shape push-button AI deployment.
Tell us what you’re building and where you want to run it.
Your model. Your hardware.
Our next conversation.
We’re inviting engineering teams to apply for the Vertis closed alpha. We’ll review your application to understand your use case and fit with the supported workflow.
01 / YOUR TEAM02 / YOUR DEPLOYMENT