04 / 06
.
AI-powered search engine
- Category
- LLM · Search
- Role
- Solo, end to end
- Link
- Visit live site

By the numbers
- 95%
- cache hit rate
- −60%
- query latency
- λ
- serverless on AWS
Overview
Built alone, Exa retrieves data in real time and streams concise, semantic answers through the OpenAI and Anthropic APIs. Most of the work went into making it fast and cheap: aggressive cache optimisation, serverless deployment on AWS Lambda and tuned PostgreSQL queries.
Highlights
- 01Streaming responses via the OpenAI and Anthropic APIs.
- 02Semantic results instead of plain keyword matching.
- 03Serverless deployment on AWS Lambda.
- 04Optimised PostgreSQL queries and caching layer.