I have been using Deepset AI Platform for around six months. I use Deepset AI Platform mainly for Gen AI model evaluation for our RAG application. For example, we are using Deepset Haystack open-source platform with a Gen AI evaluation framework called Ragas. We use Haystack to quickly assemble our RAG application pipeline to take the prompts and then interact with our vector database, which is Pinecone, and then process the query, get the response, and then compare it with the reference results provided by humans. We then use the LLM as a judge to perform evaluation and output the score for developers to see in order to evaluate the chunking strategy of their vector database.

The best feature Deepset AI Platform offers is the pipeline feature that is very easy for me to compose the large language model as well as the vector database search and retrieval, allowing me to build the application and the evaluation script within a very short period of time.

The pipeline feature and the ease of composing with large language models and vector search save me a lot of time by not writing the code from scratch. I just build the pipeline because Deepset AI Platform provides the out-of-the-box integration with the tools and stack that I am using, including the OpenAI model as well as the Pinecone API. I do not need to implement the details; I just use the existing tools in Haystack, pulling it together for the pipeline. This allows me to avoid too much detailed coding and saves me a lot of work, enabling me to focus on the evaluation.

Deepset AI Platform positively impacts our organization because we previously did not use any framework for Gen AI applications, and the introduction of this stack provides a framework for our team. It lets our team think about it and shows that it is worth introducing a framework in the future.

No improvements are needed for Deepset AI Platform.

I have been using Deepset AI Platform for around six months.

I do not have anything else to add about my main use case or how I use Deepset AI Platform in my process. I do not want to add anything else about the features. My advice for others looking into using Deepset AI Platform would be to take a look at the documentation to understand how to build the pipeline, as well as the kind of components that are provided out of the box, including the model provider and the vector database provider. Taking a look at the examples and the documentation will help to gain more insight into how to better use Deepset AI Platform. I would rate this product an eight out of ten.