Factory builds powerful scalability with MongoDB

INTRODUCTION

Factory is an enterprise software development platform that uses advanced AI agents called Droids to accelerate software development lifecycles for organizations. 

It combines agentic intelligence and state-of-the-art retrieval to help teams understand, code, review, test, and document software efficiently.

THE CHALLENGE

Constantly building with an eye to the future

From the start, Factory needed a database platform capable of handling the unique requirements of agent-native development. Its data was varied and highly unstructured, which necessitated flexible data models, robust storage, highly efficient vector search capabilities, and a powerful embedding tool.

“With AI applications, much of the data is unstructured and the applications evolve really quickly,” said Eno Reyes, Co-founder and Chief Technology Officer at Factory. “That means we needed two things: a document database that maps well with variable data structures, and a database with the flexibility to adapt as the technology changes and our product evolves.”

Upon launching, Factory switched between a variety of databases that each met business needs like scalability, reliability, or cost—but not everything at once. For Reyes, the situation was far from ideal.

“We started out with a combination of Firebase, PostgreSQL, and S3, and it was quite messy,” he explained. “We wanted to consolidate into a single database. And that led us to MongoDB Atlas.”

Factory was looking for a platform that would support significant and regular application changes; the platform also needed to support exponential growth in data volume, as Factory’s business typically processes billions of tokens daily. 

The company’s primary goals were to combine document storage, vector search, and embedding models to optimize performance and developer efficiency, while enabling scalability, reliability, and cost-effectiveness. A high level of retrieval performance from large codebases was also critical for benchmarking and application success.

“What we primarily look for when we evaluate new platforms are high performance, low cost, high scalability, and tooling,” said Reyes. “Even if we come in as an early-stage company, we’re constantly building with the idea of where we’ll be six or twelve months in the future.”

OUR SOLUTION

The best quality with the best scaling

Factory’s decision to adopt MongoDB Atlas was based on several key factors, including its ability to provide both document and vector database capabilities. This flexibility enables Factory to manage unstructured and evolving datasets, and associate text, metadata, and vector embeddings without having to juggle multiple disconnected systems.

“The balance offered by MongoDB Atlas is exactly what a fast-growing startup needs,” said Reyes. “Going with an established player, but one that still has the capability to build out as technology evolves, is the best of both worlds.”

Also valuable to Factory are MongoDB Atlas’s flexible schema and support for rapidly evolving application needs. Reyes also noted that there were lower-cost options at the time, but Factory’s longer-term outlook highlighted MongoDB’s cost-efficiency.

“The cheapest option today may end up being the most expensive in the long run if you make a mistake or sacrifice quality or reliability,” he added. “If you calculate this out, you find that the best option is the one with the best quality and the ability to scale.” 

The quality of Voyage AI’s embedding models, now part of MongoDB, further differentiated and supported Factory's choice to build on MongoDB Atlas.

“We started working with Voyage before its acquisition by MongoDB and immediately saw that it outperformed other embedding models, so we made the switch,” Reyes noted. “We were very excited because we were able to use the new model in a bunch of code-retrieval benchmarks. Until that point, we’d not seen any other embedding model that even made a dent in retrieval performance. Voyage is now core to our entire platform.”

OUTCOME

A large ecosystem with a great breadth of capabilities

The combination of improved scalability and operational efficiency provided by MongoDB Atlas enhances Factory’s agility as an AI-driven operator. By consolidating both document and vector database capabilities into a single platform, bolstered by Voyage’s embedding capabilities, Factory can efficiently handle massive data volumes, evolving data models, and critical security requirements—all of which are essential for supporting rapid growth.

“We work with a lot of relational information, and when you put all this into one platform, you unlock scalability where you don’t have to maintain separate systems,” Reyes said. “That is hugely effective for our team, because we get to focus on building the applications that we actually want to build instead of dealing with databases and all the intermediate work.”

The power of MongoDB Atlas allows Factory to scale seamlessly, processing billions of tokens daily and supporting hundreds of thousands of developers without system breakdowns or performance bottlenecks. 

“MongoDB Atlas really showcases that the vector database category is critical, but it is also a component of a larger ecosystem with a great breadth of capabilities,” Reyes added. “MongoDB Atlas definitely has that scalability and reliability locked down while it builds out new feature sets.”

And with access to advanced embedding models from Voyage AI, Factory is able to achieve superior performance in code retrieval, delivering greater accuracy and reliability across its agent-native software development platform.

“With Voyage embeddings we can outperform our benchmarking and internal evaluations of our competitors,” Reyes said. “It’s an important reason why our solution outperforms the competition.” 

By consolidating systems and reducing the need for multiple specialized tools, Factory has simplified its overall software stack, reduced potential integration issues, and accelerated deployment cycles. The result is a significant enhancement of the business’s ability to deliver a more competitive, agile, and high-performance product in the enterprise AI development space.

“If you’re building an early-stage company that is going to scale very rapidly, you need a database solution that isn’t going to break under the load of a huge volume of users,” Reyes concluded. “You need a fast-moving team with a reliable solution, and there really is one option in this space—and it’s MongoDB.”