Case Study: Perplexity achieves 40% faster foundation model training with Amazon Web Services

Perplexity Accelerates Foundation Model Training by 40% with Amazon SageMaker HyperPod

Perplexity, an AI-powered answer engine with millions of monthly users, needed massive compute power to keep up as its machine learning models grew and to speed up foundation model training. The company worked with Amazon Web Services to support its backend and training needs, including using AWS infrastructure for large-scale model development.

Amazon Web Services implemented Amazon SageMaker HyperPod to enable distributed training at scale, with checkpointing and automatic recovery to reduce interruptions. The result was up to 40% faster model training for Perplexity, helping the company advance its generative AI capabilities more efficiently.

Perplexity

Aravind Srinivas

CEO and Cofounder