AI agent knowledge development platform LlamaIndex raises $19M – News

LlamaIndex, an artificial intelligence agent development platform that automates knowledge ingestion and work for enterprise data, today announced it raised $19 million in Series A funding led by Northwest Venture Partners with participation from existing investor Greylock.

The company also announced the launch and general availability of its cloud-based managed platform LlamaCloud, a software-as-a-service offering for AI knowledge management that can enhance accuracy for AI agents.

LlamaCloud can parse, extract and index large volumes of unstructured data, including PDFs, PowerPoints, images, charts and more and turn them into workable knowledge for large language models.

Founded in 2023, LlamaIndex began as an open-source project that offered developers a jumpstart for building AI agents over any data. It includes tools such as data connectors, indexes to structure that data and advanced retrieval techniques.

AI agents are powered by vast amounts of data and systems such as retrieval-augmented generation, or RAG, which enhances large language model responses by retrieving relevant information from external sources and incorporating that data into the prompt before generating responses. This system depends on highly curated data and well-structured knowledge.

“One of the most valuable use cases for large language model agents is automating all knowledge work over unstructured data,” said LlamaIndex Chief Executive Jerry Liu. “Because only fragmented tools around data connectors, storage and agent orchestration have been available, developers struggle finding the right techniques and achieving high accuracy for production-grade agents.”

LlamaCloud is available now as SaaS or in private cloud deployments and acts as an out-of-the-box solution for building RAG applications from data ingestion through agent deployment. Enterprise customers can secure their data using role-based access controls and single sign-on and protect data access by gating it to development teams and end users.

The company also offers commercially available self-serve application programming interfaces for open-source users to access its LlamaParse product, which helps companies ingest and transform unstructured data into a structured format used in RAG applications.

In less than two years, LlamaIndex’s open-source library has grown to over 3 million monthly downloads across multiple packages and has over 38,000 stars on GitHub. The company has also taken on customers including Rakuten Group Inc., The Carlyle Group Inc. and Salesforce Inc.

“LlamaCloud’s ability to efficiently parse and index our complex enterprise data has significantly bolstered RAG performance,” said Yusuke Kaji, general manager of AI for business at Rakuten. “Prior to LlamaCloud, multiple engineers needed to work on maintenance of data pipelines, but now our engineers can focus on the development and adoption of LLM applications.”

Earlier this year, LlamaIndex announced a collaboration with Nvidia Corp. on the design and release of an Nvidia AI Blueprint for a multi-agent system that researches, writes and refines any topic using agent-driven RAG. It can be used to write blog posts based on natural language queries, which can then be critiqued and modified by the end-user until the final output has been refined to their liking. The agent system can be customized based on vast amounts of documents and other unstructured data from any source, including enterprise data, making industry research writing easier for knowledge workers.

Image: Pixabay