Why Find AI logs OpenAI requests with Velvet

This week’s article features the launch of Find AI, a high-growth startup that uses Velvet to warehouse their OpenAI request logs.

Find AI aims to replace LinkedIn with a Perplexity-style search engine. On launch day, thousands of people used the product, resulting in millions of requests to OpenAI. At peak times, they sent 1,500 OpenAI requests per second.

In the week since launch, the Find AI engineering team has relied on Velvet request logs to analyze and optimize their AI features.

What we’ll cover in this article:

• The complexities of building an AI search engine.

• Managing and optimizing the costs of AI calls.

• Model comparison, evaluations, and tuning.

More from Velvet:

We’re hiring a founding engineer! Seeking an entrepreneurial full-stack engineer. Our system is built on Next.js, Typescript, Postgres, and OpenAI.

Learn more →

Read about 4 ways to optimize your AI feature after launch. OpenAI shouldn’t feel like a black box. Use your data to analyze, evaluate, and fine-tune AI.

Read more →