MongoDB Inc. today announced that it has acquired Voyage AI Inc., a startup with a series of artificial intelligence models for generating embeddings.

The terms of the deal were not disclosed. Voyage AI previously raised $28 million in funding from Snowflake Inc., Databricks Inc. and other investors.

Nasdaq-listed MongoDB provides a popular document database of the same name. Traditional relational databases require all the records they store to follow the same format. Its document model has no such limitation, which simplifies application development.

MongoDB provides its database in several editions including a managed, cloud-based version called MongoDB Atlas. The latter offering removes the need for customers to manage the infrastructure on which the database runs. It also eases other maintenance tasks, notably patching and the process of optimizing queries to lower their hardware usage.

Atlas provides features for, among other use cases, powering AI applications. The company’s acquisition of Voyage AI is designed to enhance those capabilities. 

AI models don’t process business files in their original format, but first turn them into embeddings. Those are mathematical structures that can store key details about a file and describe how it relates to the other records in a database. Embeddings are usually generated using AI models. 

Voyage AI has developed more than a half-dozen AI models for generating embeddings. The flagship algorithm in the series, voyage-3-large, debuted last month. In terms of embedding quality, it promises to outperform competing models from OpenAI and Cohere Inc. by 9.7% and 20.7%, respectively. 

The acquired company also offers several vertical-specific models. They’re optimized to generate embeddings from code files, legal documents and financial data.

Voyager AI offers its embedding generators alongside a collection of so-called rerankers. Those are AI models that reorganize search results to display the most relevant items first. When an AI application retrieves a collection of data points to answer a prompt, it uses a reranker to identify the most relevant data points.

MongoDB plans to incorporate Voyage AI’s models into MongoDB Atlas later this year. According to the company, the integration will enable developers to turn their records into embeddings quickly and save them in the database. AI applications built on Atlas will gain the ability to find the data most relevant to a given task using Voyage AI’s rerankers. 

The database company sees the integration as a way to simplify the application development workflow. Often, software teams keep AI workloads’ standard data and embeddings in two separate databases. Voyage AI’s technology will make it more practical to store both data types in Atlas, which should simplify the coding workflow. 

“Instead of implementing workarounds or managing separate systems, developers can generate high-quality embeddings from real-time operational data, store vectors, perform semantic search, and refine results — all within MongoDB,” MongoDB Chief Executive Officer Dev Ittycheria wrote in a blog post.

After integrating Voyage AI’s models into Atlas, MongoDB plans to upgrade their capabilities. The company will add support for multimodal data such as images and videos. Additionally, it intends to introduce more industry-specific features for segments such as the legal and financial sectors.

For deeper insights on the deal, check out theCUBE Research analyst Rob Strechay’s post: A Win for MongoDB as It Acquires Voyage AI. “MongoDB is positioning itself at the forefront of the AI-enabled database market by simplifying AI application development and enhancing trust in AI-driven results,” he writes. “This acquisition strengthens MongoDB’s competitive standing and signals a broader shift toward continued embedding AI functionalities directly into core data infrastructure.”

Image: MongoDB