Skymel's NeuroSplit Revamps Old GPUs

 

Skymel has launched NeuroSplit™, the AI industry’s first Adaptive Inferencing technology, which splits AI model inferencing between end-user devices and cloud GPUs. This innovative approach significantly reduces compute costs by up to 60%, allowing AI companies to utilize older, more cost-effective GPUs. NeuroSplit also improves user experience by dynamically using idle compute capacity on devices without compromising their performance.

NeuroSplit's adaptive technology supports large, complex models that can't be fully processed locally, maintaining a buffer to ensure other applications on user devices remain unaffected. This solution features an inference pipeline cache for fast processing and avoids redundant downloads of the same models across applications, optimizing resource use and storage efficiency. 

Founded by Neetu Pathak and Sushant Tripathy, Skymel has raised $525,000 in pre-seed funding. NeuroSplit is now available in preview and will enter private beta in Q3 2024.

More Information: https://www.techdogs.com/tech-news/business-wire/skymels-neurosplit-adaptive-inferencing-lets-ai-companies-run-the-latest-genai-models-even-on-older-gpus