Neo4j goes serverless, bringing graph analytics to any data source

Neo4j Inc. today announced a new serverless offering that dramatically simplifies the deployment of its graph database offering, making it easier to use with artificial intelligence applications.

Most critically, it works with any data source, and there’s no need to fiddle around with messy extract, load and transfer operations, the company said. The new offering is meant to bring the powerful capabilities of graph analytics to any user, with one of the major implications being that they’ll be able to make smarter AI applications.

Graph databases such as Neo4j are very different from traditional Structured Query Language-based data platforms such as Oracle and Microsoft SQL. Instead of storing data in tables consisting of rows and columns, it utilizes a graph structure made up of nodes, edges and properties, to represent and store information. It’s a more versatile format that makes data easier to retrieve within a single operation in most cases.

Perhaps the biggest advantage of graph databases is that they enable what’s known as “vector search,” where unstructured data such as images and handwritten notes can be represented as vector embeddings. These capture both the explicit and implicit relationships between data and any patterns that can be drawn from it. These properties make them ideal for large language models, enabling them to retrieve a much richer variety of information, enhancing their ability to reason and infer.

As Neo4j explains, graph analytics can improve AI decision-making by “uncovering hidden patterns and relationships in complex data, delivering more accurate insights with richer context than traditional analytics.”

The analyst firm Gartner Inc. said in a 2024 report that one of the main challenges in AI development is that enterprise data is “sparse and replete with gaps,” which makes it difficult to find and link important information.

“Data and analytics leaders should use graph analytics as a preferred technology in specific use cases to fill data gaps and blend data assets even when they have diverse data quality,” the report recommended.

That’s all well and good advice, but the challenge with graph analytics has always been implementing it, as the Neo4j database and similar systems are notoriously difficult to set up and use. But that’s no longer the case with the launch of today’s new serverless offering, known as Neo4J Aura Graph Analytics.

Available starting today, Neo4j Aura Graph Analytics is said to work with any kind of data source, including Oracle, Microsoft SQL, Databricks, Google BigQuery, Snowflake and Microsoft OneLake. It’s said to make graph analytics accessible to any company by removing the biggest barriers to adoption —  namely, the need for setting up ETL pipelines, the ability to write custom queries in the Cypher language, and specialized expertise in graph analytics.

So instead of spending weeks struggling to get up and running, companies can now deploy Neo4j Aura Graph Analytics on the cloud infrastructure of their choice and start collecting, organizing, analyzing and visualizing unstructured data in a matter of minutes, the company said.

Neo4j Aura Graph Analytics comes with more than 65 ready-to-use graph algorithms and is optimized for high-performance AI applications, with support for parallel workflows ensuring any app can scale in a seamless way. Under its pay-as-you-go pricing model, customers will be billed based on the processing power and storage consumed.

“By removing hurdles like complex queries, ETL and costly infrastructure setup, organizations can tap into the full power of graph analytics without needing to be graph experts,” said Neo4j Chief Product Officer Sudhir Hasbe. “The result will be better decisions on any enterprise data source, built on a deeper understanding of how everything connects.”

The company makes some big claims regarding the kind of performance boost its new service will provide to the average AI application. Among other things, it says it can boost the accuracy of LLMs by up to 80% by helping them to uncover deeper patterns and relationships in complex connected data. Moreover, those models will be able to adapt in real time as the underlying data itself changes.

By using graph analytics, AI models can derive insights from their underlying datasets twice as fast as before, thanks to Neo4J’s use of parallelized in-memory processing. It also reduces coding tasks by up to 75%, as there’s no need for any ETL. Finally, because the offering is serverless, there’s no need to worry about the administrative overheads, which can translate to a reduced total cost of ownership as it eliminates the need to provision and maintain servers.

International Data Corp. analyst Devin Pratt said the launch of Neo4j’s serverless platform is an “exciting move” by the company that will significantly boost the accessibility of graph analytics.

“It will allow enterprises to scale analytics across any data source or cloud platform, transforming their data into a wealth of actionable knowledge, providing deeper insights for improved organizational decision-making,” he said.

Neo4j said its serverless offering will soon be joined in general availability by its native integration with Snowflake, which was first announced last year. With that integration, Snowflake users will be able to employ more than 65 graph algorithms directly, without needing to move information from that cloud data warehouse environment first.

Image: SiliconANGLE/Meta AI