Got It AI Introduces AutoRAG 2.0 for Generation of No-Hallucination, Enterprise RAG Applications With Quick Time-to-Value and Response Accuracy

Get 99.85% Accuracy for Retrieval Augmented Generation (RAG) on Enterprise Documents With an Automatically Generated, No-Code Conversational Q&A Application in Minutes

Got It AI

Got It AI, a pioneer in Gen AI RAG automation, is proud to launch AutoRAG 2.0, a game-changing Gen AI automation platform that empowers enterprises to quickly deploy trusted RAG-based applications and agents that leverage an enterprise’s internal proprietary data and knowledge base. Enterprises simply provide their LLM of choice (GPT-4, LLama3, Mixtral, etc.) and their enterprise documents or knowledge base, and AutoRAG 2.0 automatically produces a robust and accurate end-to-end optimized RAG application, ideal for improving the productivity of contact center, customer service and sales teams.

“Only four percent of Gen AI pilots moved to production deployments last year, with implementation time and hallucinations highlighted as major blockers. We’ve put a tremendous number of automation-focused and accuracy-focused innovations using our GEMS Generative MetaData System, RAG Automation, and our industry-leading TruthChecker Model into AutoRAG 2.0 — all without fine-tuning LLMs on enterprise-specific data. Through this we’ve achieved a robust product while the rest of the industry is still working on stringing individual components and models. AutoRAG 2.0 is now being rolled out by our customers,” said David Chu, co-founder. “Even 98 or 99% response accuracy is not good enough — our goal is 99.9%, which is virtually hallucination-free, for Copilots, before any Active Learning is done.”

Already deployed with multiple enterprises as a better, faster time-to-value alternative to applications developed by solutions teams or GSIs, AutoRAG 2.0 leverages three proprietary technologies developed over three years to deliver superior deployment speed and end-results: 

  • GEMS: Innovative Metadata-augmented, Graph-driven ETL to greatly improve response accuracy over a large number of complex documents with unstructured and semi-structured data.
     
  • RAG Automation: No-code generation of a full RAG pipeline with task-specific fine-tuning recipes, synthetic data generation and model evaluation tuned with experienced human evaluators. Can leverage any vector database or customer-selectable LLM. The RAG LLM is not fine-tuned out of the box on the enterprise’s data, but an enterprise may choose to further fine-tune the LLM.
     
  • TruthChecker LM: Unique technology using a low-latency and low-cost, fine-tuned, task-specific Language Model (LM) that virtually eliminates LLM hallucinations in the generated RAG pipeline. When used with GPT-4 Turbo-class models, it delivers better than 99.8% response accuracy. Even with open-source models like Llama 3.1-70B, 99.3% accuracy is achieved without fine-tuning the TruthChecker on enterprise data. Visit https://www.app.got-it.ai/blog for more information.

Enterprise IT teams are often challenged with stringing together open-source components, a vector database, large language models, fine-tuning recipes, an orchestration engine, and ETL pipelines. Due to the extensive experimentation required to put together a production application, those teams struggle to build accurate, robust RAG applications. AutoRAG 2.0 can automatically generate an enterprise-quality RAG pipeline that combines proven and sane default configurations, with innovations like GEMS and TruthChecker LM to produce extremely accurate, ready-to-deploy RAG applications. These applications can utilize cloud and/or privately hosted LLMs, and automatically benefit from hardware acceleration on industry-standard compute hardware using NVIDIA’s TRT-LLM and Triton inference server technology. 

For more information on AutoRAG 2.0, visit here: https://www.app.got-it.ai/autorag.

“We have seen enterprise IT teams spend many person-years of engineering and data annotation efforts using open-source models and components to build RAG applications, only to fall short of expectations due to unacceptable accuracy and response quality. Moreover, other companies have not even begun deploying RAG applications due to fears of ballooning development times and cost, impacting ROI. Got It AI has invested three years building a RAG product capable of generating RAG applications producing only relevant and grounded responses. This technology has been validated across thousands of users,” said Amol Kelkar, co-founder. “Having achieved this level of accuracy, we plan to now enable multi-modal and agentic RAG, which requires extreme accuracy before it can be trusted to accomplish tasks.”

To learn more about Got It AI and its innovative RAG AI platform, please visit www.got-it.ai.

Source: Got It AI