Design around real device constraints
Explore sparse model structures and execution paths that meet task-quality requirements within latency, memory and power budgets. Real-time perception, local inference and edge decision-making are key applications for this product line.
Collaborate around a specific task
Start with the target device and workload to define model adaptation, test environments and acceptance criteria. On-device continual learning remains an active research direction; capabilities and deliverables are established for each project.