SearchUnify Virtual Assistant is set to change the face of customer and employee support with its ability to deliver 24/7 personalized and contextual responses.

SearchUnify is pleased to announce that SearchUnify Virtual Assistant (SUVA) is now the world’s first federated retrieval augmented chatbot, to enable contextual and intent-driven conversational experiences at scale.

Powered with large language models (LLMs), SUVA leverages machine learning, NLP, NLQA, and retrieval augment generation to resolve customer and employee support queries in the most contextual, personalized, secure, and intent-driven manner.

“Organizations are in a race to adopt LLMs. And while, they stand to gain a lot of productivity improvements through LLMs, when a user question is directly sent to an open-source LLM-fueled chatbot, there is an increased potential for hallucinated, contextually weak responses, considering the generic dataset training of the LLM bot,” said Vishal Sharma, CTO, SearchUnify.

Marketing Technology News: MarTech Interview with Rich Donahue, Chief Marketing Officer at Ibotta

“I am super excited to share that SUVA, with its FRAG (Federated Retrieval Augmented Generation) approach, is now the world’s first and only chatbot that delivers on the organizations’ promise of contextual , connected, and personalized conversational experiences for their customers and employees in a secure, cost-effective manner,” Vishal added.

“The global chatbot market is projected to grow to over $994 million by 2024. With our future-ready SUVA, we are well-poised to capitalize on this opportunity. We are thrilled about this launch and are looking forward to helping our customers ride this innovation wave while achieving higher ROI and improved efficiencies,” said Alok Ramsisaria, CEO, Grazitti Interactive (SearchUnify’s parent company).

Marketing Technology News: The Golden Age of Immersive Media is Here

Key Differentiators of SUVA:

– Ease of setup and plug and play integration with leading LLMs, including BARD, Open AITM, open-source models hosted on Hugging FaceTM or our in-house inference models including Falcon, Mosaic, and more
– More control over response humanization with temperature control to adjust the randomness and creativity of the responses generated by the model
– Efficient costing with respect to LLM usage with segregation of intents into LLM vs non-LLM directed intents
– Better intent recognition assistance for tree-based conversation flows to ensure more flexible and adaptive conversations
– User level personalization and access controls, allowing organizations to define user roles and associate them with specific access privileges when it comes to responses
– Fallback response generation to enable seamless user experience in case of LLM downtime