NPUs are rapidly becoming a core component of modern heterogeneous SoCs, alongside CPUs and GPUs—unlocking significant performance gains for edge AI. However, as hardware evolves, software integration remains a key challenge, with developers often lacking unified frameworks to efficiently leverage all available compute resources.
In this upcoming EDGE AI Talk, we explore a flexible deployment SDK built on open-source ecosystems that enables fast and efficient integration of NPUs, supporting a wide range of AI model types. Through a real-world case study using NXP’s i.MX 95 SoC, we demonstrate how NPU support for LLM workloads was achieved in under four months.
The session will also feature an end-user perspective, showcasing how this approach enables seamless access to heterogeneous acceleration in production edge AI systems.