Lightley.ai's innovative data curation tool faced substantial roadblocks due to the inherent limitations of its GPU-based deployments. Key challenges included poor scalability, where full GPU allocation proved inefficient during periods of low demand, significant processing overhead between CPU pre-processing and GPU inference tasks, and a complex, costly procurement process for compute resources across both on-premises and cloud environments. These issues not only hindered efficient resource utilization but also burdened developer time, highlighting an urgent need for a more agile and cost-effective solution. This case study explores how Lightley.ai strategically partnered with Ampere, migrating their data curation tool to GCP T2A instances powered by Ampere Altra processors, to overcome these obstacles and fundamentally transform their AI deployments and operational efficiency.

Lightly.ai develops a cutting-edge cloud-based and on-premises data curation tool that significantly enhances the quality of video and image datasets for machine learning model training. They provide a comprehensive AI data preparation feedback loop, encompassing active learning, data exploration, online selection, and model monitoring. Computer Vision Suite Lightly.ai empowers companies to train machine learning models efficiently with optimal datasets, enabling production-ready and enterprise-grade AI solutions that process large volumes of data effectively.