Whenever you're ready, walk me through how you'd approach this.
Explain your experience with hardware abstraction layers and how you design them for scalability.
Discuss how you would define stable interfaces across heterogeneous sensors, actuators, compute devices, and ML inference backends. Address versioning, capability discovery, simulation, testing, latency, failure handling, and how the abstraction avoids leaking hardware-specific behavior into models or application code.
Provide a concise design and representative Python implementation.