Inspiration
Many AI services carry hidden x86 or CUDA-only assumptions. Before moving inference workloads to Arm64 cloud infrastructure, teams need a quick way to see what will block migration.
What it does
Arm AI Portability Auditor scans AI/ML repos and produces Markdown and JSON reports. It flags amd64/x86 platform pins, CUDA/NVIDIA-only paths, Intel/MKL/AVX/SSE assumptions, architecture-sensitive packages, x86-shaped cloud instances, and missing Arm64 validation notes. Reports include severity counts, a portability score, line-level evidence, skipped-file counts, and next actions.
How it was built
This is an AI-assisted, human-directed open-source utility for Track 2: Cloud AI. It uses only Python's standard library, includes risky and Arm-ready sample projects, regression tests, an MIT license, and a GitHub Actions workflow on GitHub's hosted Ubuntu Arm64 runner.
Validation
The risky sample returns 6 findings and score 24. The Arm-ready sample returns 0 findings and score 100. Public Arm64 validation passed here: https://github.com/copperwisp/arm-ai-portability-auditor/actions/runs/27428421782
What's next
Add ecosystem-specific rules, CI annotations, and hooks that connect static findings to real Arm cloud benchmark runs.
Demo walkthrough
A safe animated walkthrough is embedded in the README: https://github.com/copperwisp/arm-ai-portability-auditor#demo-walkthrough
Log in or sign up for Devpost to join the conversation.