Dropbox has announced that its new machine-learning based full-text search engine is ready for prime time, replacing its three-year-old Firefly engine. The new technology, dubbed Nautilus, has been designed to address the challenges faced by Dropbox in trying to build a reliable and personalized full-text solution as the company’s systems continue to scale upward in size.
As Dropbox notes, with over 500 million users that all store unique documents and have different preferences and search behaviours, the company’s needs differ greatly from traditional web search engines. With the new search engine, Dropbox is building on its Machine Intelligence Platform to work out a system that will gradually improve over time as new data is fed into it, with the objectives of delivering “best-in-class performance” and reliability, being able to create a foundation that will provide the basis for intelligent document ranking and retrieval, as well as keeping the system open-ended enough to allow Dropbox engineers to iterate and experiment on it, while still performing well and protecting user privacy.
While Dropbox has ben running Nautilus in “shadow mode” for testing purposes, the company has now switched it over to become the primary search engine at Dropbox, and reports that they’ve already seen “significant improvements” in indexing time, and expects to build more onto the platform to add new features and improve search quality. A full look at the architecture of the new platform can be found in the company’s blog post.