vAiStrata

Product

Components

Your training stack stays as it is. The runtime takes the trained model to any GPU with a Vulkan driver: an inference engine and a multi-sensor zero-copy pipeline.

Runtime Layer

The vAi and vStream pipelineCamera, LiDAR and radar input enters vStream, a zero-copy pipeline of sensor, preprocess, inference, postprocess and output stages. The inference stage is vAi, the Vulkan AI engine, built from NeuralNet, Tensor, Node and Loader modules. AI model input uses SafeTensors weights; ONNX import is planned. Output leaves as display, motor control and data stream. The same codebase runs on NVIDIA, AMD, Qualcomm, ARM and Intel GPUs.AI ModelStandard model fileSafeTensorsONNX (planned)Default onHugging Face HubCameraLiDARRadarvStreamSensorPreprocessvAiNeuralNetTensorNodeLoaderInferencePostprocessOutputDisplayMotor controlData streamSingle codebaseNVIDIAAMDQualcommARMIntel

vAi

Runs your model on Vulkan compute. One engine, every Vulkan GPU.

  • Loads SafeTensors weights in F32, F16 or BF16. The network is defined in code from the built-in operator set
  • One API across AMD, Qualcomm, Intel and NVIDIA GPUs
  • Runs on the stock driver each platform already ships, including Mesa RADV and AMDVLK on AMD. No custom driver install
  • Same interface on Linux and Windows
  • Heterogeneous execution across GPU and NPU on supported edge SoCs
Coming soon

vStream

Takes multi-sensor input end to end without letting intermediate results leave GPU memory.

  • Video decode into GPU memory
  • Preprocessing and multi-sensor fusion in place
  • Inference through vAi on the same buffers
  • Post-processing and visualization without a host round-trip
Coming soon

EVA

Easy Vulkan API. Wraps advanced Vulkan features in a smaller surface, and it is the layer vAi itself is built on.

Coming soon

Separate track

Safety certification planned

A deterministic compute driver on Vulkan SC is planned for automotive, industrial and medical deployments. Target standards include ISO 26262, IEC 61508 and ISO/PAS 8800.

Licensing

vAi and vStream are Apache 2.0. EVA is MIT licensed. There is no revenue threshold and no royalties.

About the company