Nota, a company specializing in artificial intelligence (AI) model compression and optimization technology, said Monday it will provide compression and tuning services to help enterprise customers maximize AI model performance on Amazon Web Services' (AWS) dedicated AI chip environments.
The primary customers for the service are enterprises considering the adoption of AWS Trainium, AWS's own AI training chip, and AWS Inferentia, an AI inference chip designed to deliver high-performance inference at low cost. As an AWS certified partner, Nota is launching the AI model tuning service based on its own platform, NetsPresso, which supports the entire process from AI model compression to optimal deployment for each hardware type. NetsPresso is a compression and deployment automation platform that can reduce AI model size by up to 90 percent or more while maintaining accuracy. It provides model optimization tailored to various hardware environments in a single workflow. The service consists of proof-of-concept (PoC) and diagnosis, model porting and compression, and performance tuning stages. It supports customers' AI models to operate at maximum efficiency in AWS AI chip environments.
AWS Trainium and Inferentia are proven AI chips used by global leading companies including Anthropic, Apple, Databricks, Uber, Ricoh, and Descartes. Nota's service focuses on helping customers quickly apply the performance-proven AWS AI chips to their own workloads.
Nota has previously accumulated experience tuning large language models in AWS Trainium and Inferentia environments. By applying Nota's compression technology to a high-performance language model with 32 billion parameters, the company achieved a 68 percent reduction in model size while keeping accuracy loss below 1 percent.
"AWS AI chips offer excellent price-to-performance, and Nota supports model tuning so that customers can fully utilize the performance of these AWS chips in their own models," said Nota CEO Chae Myung-soo. "We will deliver tangible cost-efficiency results to enterprise customers considering the adoption of AWS AI chip-based infrastructure."
