Traditional tabular databases weren’t made for the AI era. They don’t seem capable of handling complex data types like vectors, images, videos and audio at scale. Storing is one thing, but what about retrieving that data? If it’s one file type, then it might be easy pickings, but things get more complicated when dealing with multiple types.
LanceDB is looking to be that resource for the AI world. The Y Combinator-backed company has announced that it raised $11 million in total funding and closed an $8 million seed round to develop a database specifically fashioned for multimodal data. It’s built using the open-source Lance columnar format, a way of storing data designed for machine-learning tasks. The company boasts it has a database with native object storage integration, meaning it’s optimized for high-performance, scalable and cloud-native AI data management and retrieval.
Building Trustworth AI: Microsoft’s Strategy for Secure and Scalable Generative AI
Developers can use LanceDB in one of three ways: embed it within their existing backend, run it directly from a client application like a Jupyter Notebook, or deploy it as a remote serverless database. Unlike other databases, which run with the client and server as separate processes, LanceDB boasts that it will separate the storage from compute and can be embedded directly in apps.
Midjourney, Character.ai, Airtable, Tubi, Hex, and WeRide are some of the organizations already using the platform. With this newfound capital, LanceDB is expected to further develop its product to help developers convert their AI projects from experimentation to production. Venture firm CRV led the company’s seed round, with participation from Y Combinator, Essence VC and Swift Ventures.
“Multimodal models are the new frontier and forward-thinking AI practitioners need new wave data infrastructure and storage solutions that will permit them to train, update and keep pace with tomorrow’s AI applications,” CRV general partner Murat Bicer says in a statement.
Some of LanceDB’s current commercial offerings include its fully-featured open-source database with Rust, Python and JavaScript SDKs, a hosted serverless solution, and an enterprise product geared towards teams dealing with large datasets and requiring enterprise-grade security features.
Content Courtesy – Venture Beat
