AlphaSense was founded in 2011. The AI market intelligence company quickly scaled globally with its competitive offering: a search engine that leverages machine learning and natural language processing technology to help professionals extract insights from thousands of sources and billions of data points.
With technologies and competitors constantly evolving, businesses need a way to find and sort critical market information and insights quickly. AlphaSense’s market intelligence platform pulls from an extensive universe of public and private content, including company filings, event transcripts, news, trade journals, and equity research, so that businesses can make critical, data-informed decisions in one place.
The challenge: cumbersome data aggregation and delivery
AlphaSense aggregates data from hundreds of different sources covering a broad spectrum of clients, topics, and industries, including licensed and proprietary sources like FactSet and LexisNexis. The company uses micro frontends—a type of frontend web development architecture where an application is split into features and delivered independently—to build out features like search, dashboard, and settings. The frontend development team splits these micro frontends based on use cases and workflow, ensuring that users can see many different views within a single platform.
While this amounts to convenience for customers, revealing so many different data sources in a useful way across these micro frontends presented a technical challenge. AlphaSense’s AI capabilities rely heavily on data aggregation. In AlphaSense’s early days, the team tackled API integration, aggregation, and orchestration using typical microservices that were owned and documented by individual teams. While using a backends-for-frontends model, it would typically take their teams between days to weeks to integrate APIs and test changes to build a single feature.
“In the user management domain, we interfaced with fifteen different engineering teams, so complexity increased significantly. We needed permissions every time we wanted to update parts of a feature, and this would inevitably result in lots of back-and-forth communication to fetch, update, or navigate that feature,” said Eldar Tinjić, Engineering Manager at AlphaSense.
The solution: reducing cognitive load with the supergraph
To tackle these data aggregation issues, one of the lead architects at AlphaSense suggested GraphQL federation, which abstracted all of the complexity of the backend for frontend teams, enabling the AlphaSense team to ship faster.
“We opted to use Apollo Federation from the beginning and implemented it in production immediately,” Eldar explained. “It just came so naturally. We didn’t have to do a lot of convincing—the teams wanted to use it because it was beneficial and seemed like a step in the right direction.”
When AlphaSense first started with Apollo Federation, the company started with a few subgraphs. They set up a proof of concept in one of their development environments alongside an API gateway. The user management team was the first to use GraphQL to fetch user information for the frontend, and other teams soon began hopping on board when they saw the benefits firsthand. Now, AlphaSense has thirty-four subgraphs that vertical teams own individually. These teams receive overarching guidelines from a horizontal graph steward team.
“The subgraph boundaries are feature-driven and typically assigned to one team. Even if one team needs something from another, it usually comes as a part of dependency, and the team that owns the domain could contribute to that,” Eldar said. Soon after starting Federation, AlphaSense also began using Apollo GraphOS, which allowed them to see all operations, requests per second, breakdowns, error rates, and usage in different environments. It also granted full visibility into how schemas are built and where entities are mapped to one another.
“The biggest difference is in the release process, the testing, the regression, and the confidence that nothing has been broken. Even if contracts change, we know exactly which clients would be impacted and in what way. Before, there was no easy way to get this information, and avoiding these incidents is where we get the most productivity.”
The result:
After implementing the supergraph, Eldar estimates that AlphaSense can roll out new data sources and functionality to client teams between 20-40% faster, depending on the feature. This newfound success can be attributed to the following benefits of Apollo Federation and GraphOS:
- Reduced cognitive load for developers: By replacing backends-for-frontends and REST APIs with a self-service GraphQL platform, client and development teams at AlphaSense can work asynchronously to build features faster. The streamlined process, supported by automated workflows in the schema development process, significantly reduces the need for back-and-forth communication between teams. This efficiency has not only reduced cognitive load but has also led to a higher overall release cycle and frequency of releases, with an estimated improvement of 20-40%, thanks to the adoption of Apollo and GraphQL.
- Standardization across teams: AlphaSense now benefits from better cross-team collaboration with a central source of truth and standardized workflows. With GraphOS, development teams can enforce schema consistency across multiple subgraphs, leveraging robust schema checks that are integral to our deployment pipeline. This is especially crucial for more complex cases where modifications to existing schemas must ensure backward compatibility. The ability to configure workflows to flag inconsistent elements across subgraphs, elements, and directives further enhances our confidence in safe and seamless deployments.
- A robust ecosystem of tooling: Apollo offers an ever-growing collection of competitive tools and services to help companies like AlphaSense build their features faster and more efficiently—including OSS tools, training videos, code-gen, and other services that can be implemented immediately. Additionally, Apollo benefits from a thriving community that contributes a wealth of community-driven material, making it easier than ever for anyone to get up to speed with GraphQL and Apollo.
- Scalability: Through standardization, GraphQL allows companies to develop mastery of their workflows and policies, allowing them to scale at a much faster rate. AlphaSense benefited from improved standardization when sourcing and consuming data from the common graph.
GraphOS has allowed AlphaSense to power more AI-driven and personalized research than ever before. Now, the company can implement micro frontends more simply with a language that’s understandable for customers and engineers alike, giving them greater velocity for the backend and improved flexibility on the frontend. With the workflows that GraphOS provides, development teams at AlphaSense can rapidly deliver changes to frontends without worrying about breaking changes.
“If you’re a company that’s expecting to build more and scale more, but you still want the confidence that what you’re building is standardized and not slowing you down, then I would recommend you consider an Apollo GraphQL approach.”
Next story
Share article
