Cypris puts Elastic vector search and RAG technologies at the heart of its AI-powered research platform

Cypris is revolutionizing the research and development (R&D) sector. Its cutting-edge AI platform enables R&D teams to analyze more than 500 million technical & market level data points in seconds.

Elasticsearch, a powerful search engine that organizes and indexes vast amounts of content, sits at the heart of the Cypris platform. Once the data is retrieved, a state-of-the-art generative AI platform processes it, delivering tailored reports and dashboards in a matter of minutes. This significantly reduces the need for manual research, saving valuable time and effort.

This retrieval-augmented generation (RAG) method has been widely adopted by Cypris clients across industries, including manufacturing, defense, and pharmaceuticals. Cypris CEO Steve Hafif says, "By simply entering a few keywords, you can gain a comprehensive understanding of the latest innovations in fields like nuclear energy or electric vehicles within seconds, answering key R&D questions. Cypris extracts critical information from a wide array of technical and market-focused data points to profile organizations, identify unique trends, and predict where sectors are evolving."

Unlike traditional chatbots that hallucinate and offer limited visibility into underlying data, Cypris narrows the large language model's context window to its real-time, innovation-focused database, with continuously updated data from global patents, scientific literature, funding institutions, organizations, news, and more. This seamless integration streamlines the research process, allowing researchers to build live reports on the state of global innovation in their research areas. 

Cypris' decision to adopt Elasticsearch was driven by its advanced capabilities and commitment to customer success. Elastic's tooling around semantic search inference pipelines and dense vector queries proved useful for their text-based search applications compared to other search solutions. "Having the option to search through a large corpus of documents with BM25 and/or vector similarity ultimately led to more complete and relevant responses," Says Principal Engineer Logan Pashby who spearheaded the search and RAG projects.

Having a native vector database in Elastic helped them quickly go from zero to one on semantic search. They chose the dense vector approach to utilize a model that encodes a rich representation of their data. Support for third-party model weights enables Cypris to fine-tune search while continuously growing with Elastic as the database improves.

The hybrid search capabilities make Elastic even more powerful. Being able to score relevance as a combination of vector similarity and traditional Elastic queries like multi-match, filtering, and fuzziness empowered Cypris to support their more niche search use cases.

Beyond its technical capabilities, their team was impressed by Elastic's proactive support and commitment to collaboration. "Our successful implementation of semantic search was sped up by the expertise of Elastic's team," he says. "Their willingness to provide insight, even down to the lowest level of the dense vector search execution, demonstrated their dedication to our success."

Elastic's reliability and scalability were crucial in Cypris' decision to adopt Elasticsearch. After encountering scalability problems with their previous search provider, Elasticsearch proved to be a game-changer. Timeouts and cluster failures are now a thing of the past, even during peak usage. With Elastic, Cypris can effortlessly scale its search infrastructure without technological constraints. This robust scalability has allowed the company to manage over 500 million documents, totaling more than 10 terabytes of data.

Cypris' commitment to security has been instrumental in securing high-profile clients within the U.S. Department of Energy and Department of Defense. These organizations conduct rigorous security audits, scrutinizing every aspect of Cypris' systems, including those that use Elastic.

Hafif emphasizes Elastic's exceptional security record. "There has never been an issue with Elastic. Their strong security practices and U.S.-based operations provide us with the utmost confidence when winning government contracts," he says.

Beyond security features, Elasticsearch has significantly enhanced Cypris' commercial appeal, driving quarterly customer growth of nearly 30% and attracting multi-million-dollar investments from venture funding partners. "Our business metrics are ultimately driven by successful, satisfied clients," says Hafif. "With Elasticsearch at the heart of our technology stack, we can continue to deliver an unrivaled research experience."

Looking to the future, Elasticsearch remains the foundation for rapid growth—both data storage and client acquisition. Cypris anticipates surpassing a billion stored documents by the end of the next year through growing their data partnerships and expanding their existing core database, all of which will be running through Elastic.