Contextual AI Inc., one of the leading players in retrieval-augmented generation systems that can advance the capabilities of large language models, today announced the general availability of its flagship product.

With the launch of the Contextual AI Platform, enterprises will be able to create specialized “RAG agents” that can automate knowledge work on behalf of users.

The new platform is based on an improved version of the company’s technology, known as RAG 2.0, which debuted last year. The company says the technology enhances the way AI models can retrieve data from third-party datasets to improve their knowledge, improving the accuracy of their outputs.

The Contextual AI Platform provides access to all three of the main components needed to build a RAG system, including the underlying LLM that responds to questions, a “retriever” module that enables the LLM to query external data sources, and an embedding model that can transform raw data such as images into mathematical structures that can be processed more easily.

By providing all three components, Contextual AI says it simplifies the development of RAG systems, saving time and effort on their integration.

The company is led by its Chief Executive Douwe Kiela, who was one of the pioneers of RAG, helping to build the first such system in 2020 while working with Meta Platforms Inc.’s AI research team. Prior to that discovery, LLMs could only leverage the information that existed in their original training datasets. With RAG, they can access external data, vastly improving their usefulness.

Most AI practitioners agree that RAG will be essential for the creation of AI agents, which are more advanced models that can perform tasks on behalf of users with minimal supervision. Contextual AI is focused on helping companies to build “specialized RAG agents” that it says will be able to perform more sophisticated, domain-specific knowledge work on a par with human subject matter experts.

Those agents will be able to intelligently orchestrate data retrieval and generation based on the context of their user’s inputs and prompts, delivering more accurate responses for complex knowledge tasks, the company said.

Kiela said enterprise AI is reaching a critical turning point, with AI agents set to become available to the employees of every major company.

“However, the specialized work of subject-matter experts remains largely underserved,” Kiela said. “Specialized RAG agents built on the Contextual AI Platform will bridge this gap, enabling small and medium-sized enterprises to boost their productivity with AI that truly understands their domain.”

To back up these claims, Contextual AI has released its latest public benchmark results, which reveal an impressive, industry-leading performance for RAG workflows across a range of enterprise domains. Its specialized RAG agents easily surpassed the performance of well-known frontier models like OpenAI’s GPT-4o and Anthropic PBC’s Claude 3.5 Sonnet in areas such as document understanding, reranking, groundedness and overall accuracy.

Contextual AI says its platform is already being used by a number of Fortune 500 companies, including the smartphone chipmaker Qualcomm Inc.

“Contextual AI gives me confidence that we can leverage generative AI to support our team, help our customers design and develop products efficiently, and set new standards for performance and quality,” said Yogi Chiniga, Qualcomm’s vice president of customer engineering.

Image: SiliconANGLE/Microsoft Designer