Product
Components
An inference engine and a pipeline built on the same Vulkan compute foundation, so a workload written once runs on the GPUs a deployment actually has.
vAi Vulkan AI Inference Engine
Model execution on Vulkan compute rather than a vendor-specific compute API.
- Single API surface across AMD, Qualcomm, Intel, and NVIDIA GPUs
- Uses each platform's stock driver as shipped — 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
vStream Zero-Copy GPU Pipeline
End-to-end pipeline for multi-sensor input where intermediate results never 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
EVA Vulkan Wrapper
Easy Vulkan Api — making advanced Vulkan features easy to use. Published under MIT, separately from the two above.
Separate track
Safety certification planned
A separate deterministic compute driver on Vulkan SC is planned for automotive, industrial, and medical deployments. It is its own piece of work rather than a mode of the components above. Target standards include ISO 26262, IEC 61508, and ISO/PAS 8800. This work has not shipped.
Licensing
vAi and vStream are source-available. Free for organizations under $10M in annual revenue; a commercial license applies above that threshold.
EVA is MIT licensed and carries no revenue condition.