If you're preparing training data for LLMs at any serious scale, this wraps NVIDIA's NeMo Curator toolkit for Claude Code. The main win is GPU-accelerated fuzzy deduplication that runs 16 times faster than CPU alternatives, which matters when you're processing terabytes of web scrapes or Common Crawl data. It also handles content filtering and scales across GPU clusters with near-linear performance. The 40% lower TCO claim is notable if you're currently burning through CPU hours for data curation. This is purpose-built infrastructure tooling, not a general-purpose utility, so you'll know immediately if you need it or not.
npx skills add https://github.com/davila7/claude-code-templates --skill nemo-curator