![](https://d15shllkswkct0.cloudfront.net/wp-content/blogs.dir/1/files/2024/08/Ragie.png)
![](https://d15shllkswkct0.cloudfront.net/wp-content/blogs.dir/1/files/2024/08/Ragie.png)
Ragie Corp., a new startup that helps developers build retrieval-augmented generation applications, launched today with $5.5 million in seed funding.
Craft Ventures, Saga VC, Chapter One and Valor provided the capital.
Retrieval-augmented generation, or RAG, is a machine learning technique that makes it easier for large language models to learn new information. In the past, expanding an LLM’s knowledge repository required retraining it, which can be a highly expensive process. RAG makes it possible to make additional data available to a model without retraining it.
But though the technology simplifies artificial intelligence projects in certain respects, implementing it still involves a significant amount of work. As a result, building RAG-enabled applications can take months in some cases. San Francisco-based Ragie has developed a cloud platform that promises to streamline the workflow.
Ragie allows RAG-enabled LLMs to draw on data stored in popular cloud applications such as Salesforce, Google Drive and Notion. The company says that connecting new data sources takes only a few clicks with its platform. Ragie monitors for changes to the datasets it ingests and automatically makes updates available to the LLM that is using the information.
The company’s platform doesn’t stream information to AI models in its original form. Ragie first turns the data into embeddings, mathematical structures that LLMs use to store knowledge. The platform keeps embeddings in a cloud-based vector database optimized to hold such files.
Ragie also provides a so-called chunking tool as part of its feature set. Chunking is the process of splitting up the documents that an RAG-enabled LLM uses to generate prompt responses into smaller, more manageable files. Reorganizing datasets in this manner can improve the quality of AI models’ output.
Another prerequisite to building a RAG application is a reranking algorithm. Such algorithms analyze the documents that an LLM uses to answer a question and prioritize them by relevance, which ensures that only the most pertinent information finds its way into the model’s answer. Ragie’s platform provides a prepackaged reranking feature to save time for developers.
The platform also offers a number of more specialized capabilities. One feature makes it easier to build AI chatbots that can fetch information from not one but multiple sources in response to user questions. Another capability improves RAG applications’ ability to perform data extraction tasks such as identifying all the earnings reports in a large collection of business documents.
The company announced its funding round in conjunction with the launch of its platform into general availability. Ragie provides a free tier that allows developers to process up to 10 RAG requests per minute at no charge. The company generates revenue through two paid editions of the platform, Pro and Enterprise, that offer higher rate limits.
THANK YOU