Technology · AREN runtime
Any model, any silicon, automotive‑grade.
AREN is our automotive AI runtime — the execution layer that runs learned workloads on the vehicle’s silicon, keeps them updated, and manages their lifecycle for the life of the car. Neural audio is the first workload shipped on it.
What the claim means, concretely.
A supported-model contract
Workloads target AREN, not a specific chip. A model that meets the contract — operators, precision, memory, scheduling behavior — runs wherever AREN runs, and its authors stop caring which vendor's NPU sits underneath.
Heterogeneous silicon targets
Vehicles ship with whatever silicon the program negotiated, and it changes mid-cycle. AREN treats NPUs as interchangeable execution targets behind one interface, so a workload validated once is not re-engineered per SoC.
Deterministic integration
Learned workloads get budgets, not promises: bounded latency, bounded memory, partitioning away from safety functions, and a fail-safe path that no model output can bypass. Designed to functional-safety practice from the start.
Model update, over the air
A learned model is a versioned asset with a lifecycle — deployed, monitored, rolled back, improved — across a fleet and across years. AREN owns that lifecycle on the vehicle side.
Engineering principles.
These are the properties an OEM review can hold us to. None of them is a certification claim.
- Deterministic executionfixed budgets, sample-rate paths in deterministic code
- Partitioninglearned workloads isolated from safety functions
- Fail-safe limitera hard stage no model output can bypass
- Functional-safety practicedesigned to it — claimed only when held
Where it stands.
One honest sentence about maturity: AREN is under active development within national software-defined-vehicle research programs, and it is proven first by our own neural audio workloads — the same discipline we apply to every claim on this site.
The audio line is the existence proof. The neural audio page describes the first workload AREN carries, at the level of detail we can defend today.