Quick Start Guide for iMware Starter
iMware is a comprehensive software platform for developing and optimizing AI applications on intellino NPUs.
iMware is an end-to-end software platform designed to help developers build, analyze, and deploy AI workloads efficiently on intellino NPU products. It streamlines the full development cycle—from early performance exploration to final deployment—by combining workload analysis, specification-aware performance evaluation, and a robust set of SDKs and libraries. imWare offers two editions to support users from first-time NPU development to advanced architecture design.
imWare Starter: Start building AI on NPU, simply and fast.
imWare Architect: Analyze, optimize, and architect AI on NPU.
When implementing AI applications on an intellino NPU, understanding how a model behaves under real system constraints is critical. iMware supports:
Workload profiling and characterization to identify compute, memory, and bandwidth bottlenecks
Performance analysis based on detailed product specifications, enabling realistic estimation and comparison across intellino NPU configurations
Layer-by-layer and operator-level insights to pinpoint inefficiencies and guide optimization decisions
Latency, throughput, and utilization analysis to support both real-time and batch inference scenarios
These capabilities help teams make faster architectural decisions and reduce iteration cycles during product development.
iMware provides a unified SDK that enables developers to move from models to deployable binaries with minimal friction. The SDK is designed to support:
Model import and conversion from common AI frameworks
Compilation and optimization aligned with intellino NPU execution characteristics
Runtime APIs for application integration and system control
Deployment workflows for embedded and edge environments
This integrated approach accelerates development while maintaining control over performance and resource usage.
To shorten development time and improve portability, iMware includes optimized libraries commonly required in production AI systems, such as:
Pre-/post-processing libraries for vision, audio, and sensor pipelines
Operator and kernel libraries optimized for intellino NPU execution
Utility libraries for memory management, scheduling, and profiling
These components allow developers to focus on application logic while relying on tuned building blocks for performance-critical paths.
iMware is designed not only for application deployment, but also for intellino NPU product development and validation, supporting:
Early-stage feasibility analysis against target specifications
Performance regression tracking across hardware and software revisions
Repeatable benchmarking for product tuning and release readiness
Performance estimation and optimization for intellino NPU–based AI applications
Workload-driven selection of intellino NPU product configurations
Rapid development using SDKs and optimized libraries
Embedded and edge deployment requiring deterministic performance and efficiency
In summary, iMware combines workload analysis, specification-aware performance evaluation, and a production-ready SDK and library stack to enable faster, more confident development of high-performance AI applications on intellino NPU platforms.