Windows ML and NPU Acceleration - Building Smarter Apps

Hands-on lab notes from LAB584-R1: Windows ML and NPU Acceleration.

Lab Overview

What We're Building: Build a smarter image classification app with Windows ML, now generally available. This hands-on lab covers:

• Dynamically downloading execution providers (EPs) for NPUs
• Compiling models for hardware-specific EPs
• Running inference locally with optimized performance
• Deploying, debugging, and optimizing using WinML APIs

Technologies:

• Windows ML (generally available)
• NPU (Neural Processing Unit) acceleration
• Execution Providers (EPs)
• Hardware-specific model compilation
• Local inference optimization

Key Learning Goals

1. Windows ML Architecture - How does Windows ML enable on-device AI?
2. NPU Acceleration - What performance gains do NPUs provide?
3. Execution Providers - How do EPs enable hardware-specific optimization?
4. Model Compilation - How do you compile models for specific hardware?
5. Deployment - How do you deploy and debug WinML apps?

Stay Tuned

Full lab walkthrough, code samples, and performance optimization guide coming soon.

Session: LAB584-R1 | Nov 20, 2025 | Moscone South, Level 3, Room 308