Case Study: Perplexity achieves a more accurate, responsible conversational search experience with Amazon Web Services

Perplexity Builds Advanced Search Engine Using Anthropic’s Claude 3 in Amazon Bedrock

Perplexity, the conversational search engine company, needed a way to deliver accurate, helpful, and safe answers while reducing hallucinations in large language model outputs. To support this challenge, Perplexity used Amazon Web Services, specifically Amazon Bedrock with Anthropic’s Claude 3, as part of its search experience.

Amazon Web Services helped Perplexity build a more responsible and accurate search engine, with Claude 3 enabling concise natural-language responses, long-document analysis, and stronger accuracy than earlier models. The results included reduced hallucinations, improved developer experience, and access to a 200,000-token context window; separately, Perplexity also reported that AWS-powered training infrastructure doubled throughput and significantly reduced hardware failures.

Perplexity

Aarash Heydari

Cloud Infrastructure Engineer