Image Source: Unsplash

The increased adoption of AI has led to a surge of new use cases across all industries. One such interesting use-case is being developed by San Francisco-based Exa which is trying build a search engine for AI.

Exa’s Offerings

Exa was founded in 2021 by Harvard alumni and good friends, Will Bryk and Jeff Wang. Launched even before ChatGPT, Exa was initially set up with a goal of using AI to build better search. Soon after ChatGPT’s release though, AI companies wanted to embed Exa’s search engine into their models to use it as a search engine for their AI models.

Exa’s search engine has been designed for AI models to help these models search for information on the Internet and return accurate answers. It uses vector databases and trains models to predict the next relevant link instead of the next relevant word in the search query. Its model allows Exa to process link datasets and provide unique search results.

Unlike traditional search models that search using keywords, Exa has developed its search engine using end-to-end neural networks that get better with more experience. By using embedded models that convert web page content into a list of values, Exa can deliver more accurate search results.

Exa has built a community of thousands of developers along with AI companies, such as Databricks, as its customers. Its engine allows its customers to have complete control over their searches by providing for filters such as date, category, and domain. Developers can choose if they want to get just links, full parsed text content, key highlights or even customizable summaries per URL. Exa allows users to search using natural language queries, making it more intuitive and user-friendly. By offering low latency and high reliability, the engine has been built to allow to scale up as needed. Interestingly, Exa’s engine was used by Databricks to find large training sets for its own model training initiatives.

Exa’s Financials

Exa operates on a multiple tiered-fee pricing level. It offers a free tier that allows anyone to try its search engine. Tiers start at $5 per 1,000 requests that deliver 1-25 results from its neural model. It also has an enterprise pricing structure for its bigger customers. Exa does not reveal its revenue, but has said that they are generating revenues, and the revenue is rising.

Exa has raised $22 million in funding so far. Its last round was held in July last year when it raised $17 million in a round led by Lightspeed, Nvidia’s NVentures, and Y Combinator. Valuation details were not disclosed. Exa is focused on catering to AI-driven applications. It is not competing with Google. Its unique approach to search is helping AI companies evolve themselves. It will be interesting to see how Exa continues to grow in the future and adapts to the evolution of AI.

More By This Author:

Duolingo Empowers Lily With AI
Cloud Stocks: MongoDB Focuses On AI Platform Development
Cloud Stocks: Symbotic Leverages AI To Provide Warehousing-As-A-Service