Got It AI to Drive Platform Partnerships With Infrastructure Leaders Bringing Gen AI Platform ELMAR With NVIDIA AI Solutions for On-Prem
PALO ALTO, Calif., October 12, 2023 (Newswire.com) - Got It AI, a leading innovator in Generative AI, has announced its plan to integrate NVIDIA AI Solutions into its ELMAR (Enterprise Language Model ARchitecture) platform. Introduced in March 2023, ELMAR is an on-premise enterprise AI platform with unique features like PII masking, prompt-injection defense and LLM hallucination reduction. The integration with NVIDIA AI Solutions will enhance ELMAR's manageability and performance for AI workflows, utilizing Triton, NeMo Guardrails, and TensorRT-LLM.
Krishna Hegde, a former VMware executive, has joined Got It AI to lead the company's platform partnerships. This move will also enable NVIDIA AI Enterprise partners such as VMWare, HPE and DELL to leverage ELMAR. Got It AI’s platform, offering both cloud-based and on-premise LLMs, is an ideal solution for enterprise customers seeking on-premise options, human-level accuracy, and high performance for Generative AI use cases, such as enterprise chatbots. Got It AI’s Agent Copilot and CX AutoPilot are advanced enterprise chatbot solutions that use enterprise knowledge bases to deliver reliable and integrated conversational access to enterprise data in a variety of documents, web pages, presentations, complex PDFs, and tabular information, making them ideal for sales assist and customer service applications.
For OpenAI LLMs, ELMAR’s hallucination guardrails reduce hallucinations down to human level performance, as benchmarked on hundreds of articles of real customer data.
OpenAI Hallucination Rate Test Results (model, Baseline hallucination rate*, Hallucination rate* with Got It AI hallucination guardrails):
- GPT-4, 2.3%, 1.1%
- GPT-3.5, 10.4%, 1.5%
* Hallucination rate measured on a set of 530 Q&A pairs over a 200 document knowledge base from a Got It AI customer
For Open Source LLMs, ELMAR comes with a choice of models such as Llama2 and MPT, as well as a native 3B parameter LLM which delivers the lowest hallucination rates among on-prem model options, while being substantially smaller.
On-prem Capable, Pre-Trained LLM Hallucination Rate Test Results (model, model size, Baseline hallucination rate*, Hallucination rate* with Got It AI hallucination guardrails):
- Llama 2, 13 billion, 36.8%, 21.8%
- MPT, 30 billion, 30.0%, 15.5%
- Llama 2, 70 billion, 25.4%, 12.7%
- Got it AI ELMAR, 3 billion, 17.7%, 7.5%
* Hallucination rate measured on a set of 530 Q&A pairs over a 200 document knowledge base from a Got It AI customer
The integration of NVIDIA AI Solutions into ELMAR will enhance model loading and management capabilities by utilizing the Triton Inference Server. It will also improve inference performance through the use of TensorRT-LLM. For fine-tuning, which is often required for on-premise capable models, the DGX Cloud can be leveraged as needed. Thus, ELMAR brings together NVIDIA's AI training and inference technologies into a unified solution for Enterprise companies.
Got It AI is excited to announce that Krishna Hegde, former executive managing Partner Revenue at VMware, has joined the team to spearhead strategic partnerships. Hegde brings a wealth of experience in on-premise platforms and partnerships. "I'm thrilled to apply my expertise from VMware to help elevate ELMAR in the enterprise LLM landscape," Hegde said. He praised Got It AI's efforts in integrating its optimized platform with NVIDIA’s on-premise architecture, simplifying deployments for enterprise use cases. He also highlighted the unique flexibility ELMAR offers to on-premise data center customers, allowing them to choose from a suite of open source LLMs or use a hybrid approach with OpenAI and Google cloud models, as well as allowing the customer to benchmark the model of choice for their specific enterprise data.
Peter Relan, Chairman of Got It AI, expressed his confidence in Hegde's abilities. "Krishna and I have collaborated at multiple companies, including Openfeint & Discord, where he led partnerships with industry giants like AT&T and Android app store leaders globally. He also worked at one of the first cloud-based GPU companies, Agawi, which was later acquired by Google," said Relan. He is confident that Hegde's expertise will help co-founders Amol Kelkar and David Chu in their mission to make high-performance, enterprise-class Generative AI with guardrails deployable with human level accuracy in on-premise data centers.
Got It AI is eager to forge partnerships with infrastructure and cloud providers, positioning ELMAR as the premier PaaS layer for both cloud-based and on-premise enterprise LLM use cases. For more information about Got It AI and its innovative solutions, visit www.got-it.ai.
Got It AI continues to push the boundaries of Generative AI, driving innovation and delivering high-performance solutions for enterprises with Agent Copilot and CX AutoPilot. With the integration of NVIDIA AI Solutions and the addition of Krishna Hegde to the team, the company is poised to make significant strides in the AI industry.
Triton, NeMo, TensorRT-LLM and DGX Cloud are trademarks of NVIDIA. Llama is a trademark of Meta. MPT is a trademark of Databricks. ELMAR is a trademark of GotIt! Inc.
Source: Got It AI