MongoDB said it has acquired Voyage AI in a deal that aims to make embedding and reranking of models a native part of its database.
Voyage AI embeds and reranks models based on accuracy in retrieval augmented generation (RAG). MongoDB's bet is that these tools should be in the database layer to improve accuracy of models before they wind up in the AI agent pipeline.
Terms of the deal weren't disclosed.
Here's a look at what Voyage AI does from its documentation:
In a blog post, MongoDB CEO Dev Ittycheria said Voyage AI will bring the ability to rerank results in RAG to ensure accuracy. "We believe embedding generation and reranking, as well as AI-powered search, belong in the database layer, simplifying the stack and creating a more reliable foundation for AI applications. By bringing more intelligence into the database, we help businesses mitigate hallucinations, improve trustworthiness, and unlock AI’s full potential at scale," said Ittycheria.
Here's how MongoDB sees Voyage AI working in its stack.
Voyage AI will bring the following to MongoDB:
- A strong ecosystem on Hugging Face with Anthropic, LangChain, Harvey and Replit using Voyage AI.
- Enhanced vector searches that capture meaning across text, images, PDFs and structured data.
- Improved accuracy through advanced reranking models.
- The ability to fine tune models for different industries.
As for the integration, MongoDB will continue to make Voyage AI available in AWS and Azure marketplaces. Voyage AI will be embedded into MongoDB Atlas starting with an auto-embedding service for Vector Search followed by native reranking and then industry use cases. MongoDB also is planning enhanced multi-modal tools and instruction-tuned models.
With this integration, MongoDB said developers won't need to manage external embedding APIs, standalone vector stores or complex search pipelines.
Holger Mueller, an analyst at Constellation Research, said the Voyage AI acquisition will make MongoDB a key player to curb hallucinations among models:
“Hallucination is a real AI problem, and the most relevant way of avoiding it through grounding is to query a trusted database. This is why the acquisition of Voyage AI by MongoDb is critical for MongoDB customers as it allows them in an easier and more efficient way to ground their AI models with embeddings (of code) and reranking (to find the right data) of Vector AI - all in their MongoDB data. It is critical that Vector AI remains available for 3rd party and other services as grounding demand is bigger than any single data store.”
