General Availability of MongoDB Atlas Vector Search Knowledge Bases for Amazon Bedrock

MongoDB has publicly launched its Atlas Vector Search integration with Amazon Bedrock, first showcased at Amazon Re:Invent last year. This collaboration enables developers to synchronize their foundation models and AI agents with proprietary data stored in MongoDB, enhancing the relevance, accuracy, and personalization of responses through Retrieval Augmented Generation (RAG).

"Many businesses are concerned about the accuracy of outputs from AI systems while safeguarding their proprietary data," stated Sahir Azam, MongoDB’s Chief Product Officer. "We’re simplifying the process for joint MongoDB and AWS customers to utilize various foundation models hosted in their AWS environments. This allows them to develop generative AI applications that securely incorporate their proprietary data within MongoDB Atlas, thereby improving accuracy and enhancing user experiences."

Amazon Bedrock, AWS’s managed service for generative AI, serves as a central repository for enterprise customers' AI application development needs. The growing array of models includes those from Amazon, Anthropic, Cohere, Meta, Mistral, and Stable Diffusion. While utilizing externally trained models can be beneficial, companies often prefer using their own databases for richer customer context.

MongoDB's integration is crucial in this regard. Developers can customize their chosen foundation models with proprietary data, enabling the seamless development of applications around newly-trained LLMs without manual intervention. "You can create generative AI applications, but if you can't integrate your real-time operational data, you'll receive generic responses," explained Scott Sanchez, MongoDB’s Vice President of Product Marketing and Strategy, during a press conference.

"This integration with MongoDB makes it easy for customers to connect the dots," he added. "They can privately customize their large language models with proprietary data by converting it into vector embeddings stored in MongoDB. For instance, a retailer could create a generative AI application using autonomous agents to manage tasks such as real-time inventory requests or customer returns."

This announcement follows previous collaborations between MongoDB and AWS, including MongoDB’s Vector Search available on Amazon SageMaker and Atlas support from CodeWhisperer. MongoDB continues to innovate, unveiling initiatives like its AI Applications Program (MAAP) to assist enterprise customers in developing AI applications.