We migrated from Amazon OpenSearch to MongoDB Atlas Vector Search

When we decided to adopt Generative AI technologies at Base39, we needed robust and efficient solutions to handle the enormous amount of data we analyze daily. Initially, we used Amazon OpenSearch for this process. It was a solid solution, but as our requirements grew and our data flow became more complex, we realized we needed something more agile and easy to integrate with the rest of our stack. That’s when MongoDB Atlas Vector Search came into play.

The main reason we switched from Amazon OpenSearch to MongoDB Atlas Vector Search was the simplicity and efficiency of integration with what we already use. MongoDB had already been our primary database since the beginning of Base39, so it made perfect sense to seek a vector search solution that fit directly into it. This change eliminated the need to manage yet another independent and complex tool, which significantly reduced the maintenance burden for the technical team.

Another key point was performance. With MongoDB Atlas Vector Search, we were able to accelerate the ingestion and search of vectorized data, allowing for faster and more accurate analyses. The performance gain was immediate and, together with other tools like Amazon Bedrock and automated triggers from MongoDB, we managed to orchestrate complex processes in a much smoother way.

Finally, our team’s familiarity with MongoDB was also a major differentiator. When we introduce new technologies, we want to ensure that the adaptation time is minimal so we can focus on what really matters: generating value for our clients. With MongoDB Atlas Vector Search, we were able to leverage the knowledge we already had and apply it directly, which allowed us to scale rapidly without losing efficiency.

This transition not only simplified our operation, but also allowed us to focus on delivering faster and more accurate solutions for our clients, with a technology stack that we already know deeply.