Vantar AI
Infrastructure for
physical AI.
Software is eating the physical world. Robots, edge devices, neuromorphic chips - they all need better developer tools. Vantar builds the missing infrastructure layer: from robotics CLIs to neuromorphic compilers to edge deployment.
Products
Four products, one mission. Developer tools for the physical world - from first prototype to production deployment.
Torq
LIVEThe fast robotics development toolkit
Build, simulate, deploy robots with one CLI. Written in Rust.
Nuro SDK
LIVEUniversal SNN compiler
Train on GPU, deploy to Loihi 2, SpiNNaker 2, Akida. Zero code changes.
Vantar Cloud
BETAExperiments in the cloud
Push experiments, compare runs, share recordings. Remote hardware access.
Dev Kit
EXPLORATIONEvent camera meets neuromorphic chip
Hybrid sensor + processor edge module. From sensor to inference at under 1mW.
Why Vantar
Physical AI is where software meets the real world. The tooling hasn't kept up. We are fixing that.
Robotics tooling is broken
Robotics developers spend 40-50% of their time on environment setup, dependency management, and build system wrestling. Torq replaces the entire fragmented toolchain with one CLI - init, build, sim, deploy.
Neuromorphic has no DevTools
27+ SNN frameworks exist for simulation and training. Zero tools exist for experiment tracking on neuromorphic hardware. Nuro fills the gap - record, track, visualize, and deploy across any chip.
Physical AI needs a platform
Robots, edge devices, neuromorphic chips - they all need better developer tools. From training to edge deployment, one company, one ecosystem. That is what we are building.
Torq
The fast robotics development toolkit. One CLI replaces 15 minutes of setup hell. Written in Rust.
Terminal
$ torq init my-robot --template mobile
Created my-robot/ with Nav2 mobile template
ROS2 Jazzy workspace ready
$ torq build
Cached build — 3.2s (was 4 min with colcon)
0 warnings, 0 errors
$ torq sim
Launching MuJoCo — warehouse.xml
Robot spawned. Ctrl+C to stop.
$ torq deploy
Deploying to jetson-orin@192.168.1.50
Syncing build artifacts... done
Restarting robot.service... running75x
Faster builds
Cached Rust pipeline
5
Templates
minimal, mobile, arm, drone, humanoid
3
Simulators
MuJoCo, Gazebo, Isaac
1
CLI
init, build, sim, deploy
Nuro SDK
The universal SNN compiler. Train on GPU with surrogate gradients. Deploy to neuromorphic silicon. Zero code changes.
import nuro
# Define a spiking neural network
graph = nuro.Graph()
graph.add(nuro.neurons.LIF(128, tau=20e-3))
graph.add(nuro.neurons.LIF(10, tau=20e-3))
# Train on GPU — surrogate gradients
model = nuro.compile(graph, target="gpu", requires_grad=True)
model.fit(train_data, epochs=50)
# Deploy to neuromorphic silicon — same model
loihi = nuro.compile(graph, target="loihi") # auto-quantizes
spinn = nuro.compile(graph, target="spinnaker")
akida = nuro.compile(graph, target="akida")
# Track experiments
exp = nuro.experiment("snn_baseline", project="edge_vision")
exp.set_hardware("loihi-2", board_id="ncl_04")
exp.add_recording("spikes", loihi.record())
exp.log_metrics({"accuracy": 0.94, "power": "0.8mW"})
exp.save("./experiments")5
Backends
GPU, Loihi, SpiNNaker, Akida, Samna
4
Neuron models
LIF, Izhikevich, AdEx, IF
227
Tests passing
Full stack coverage
0
Code changes
Same model, any hardware
Roadmap
Four layers. CLI to compiler to cloud to silicon. Each product is useful standalone - together they are the full stack for physical AI.
Torq CLI
Open-source Rust CLI for robotics. Project scaffolding with 5 templates, cached builds, simulator orchestration (MuJoCo, Gazebo, Isaac), and SSH deployment to robot hardware. One tool replaces the entire ROS2 setup workflow.
Nuro SDK + DevTools
Universal SNN compiler with experiment tracking. Train on GPU with surrogate gradients, deploy to Intel Loihi 2, SpiNNaker 2, or BrainChip Akida. Hardware-agnostic recording, experiment tracking, and visualization built in.
Vantar Cloud
Push experiments to the cloud. Compare runs across hardware platforms. Share recordings with collaborators. Remote access to neuromorphic chips and robot simulators you don't own yet.
Vantar Dev Kit
A complete edge AI module for event-based vision. Hybrid event camera paired with a neuromorphic processor. Nuro pre-installed. One device from sensor to inference.
Built For
Robotics Engineers
Build and deploy robots faster with Torq. One CLI replaces fragmented ROS2 tooling - scaffold projects, cached builds, simulator orchestration, and hardware deployment over SSH. Stop wrestling with setup.
Neuromorphic Researchers
Track experiments and deploy SNNs with Nuro. Record spike trains from any hardware, log parameters and metrics, visualize with one line, and compile to Loihi, SpiNNaker, or Akida with zero code changes.
Edge AI Teams
Full stack from training to silicon to device. Train on GPU, deploy to neuromorphic chips, run on edge hardware. One ecosystem covers the entire pipeline from research to production.
Join the ecosystem.
Open source. Apache 2.0. From robotics CLIs to neuromorphic compilers to edge deployment - one platform for physical AI.
Early Access - Cloud & Hardware