Lyzr Research
Inventors of AutoRAG, Automata, MagicPrompts
We continue to uncover new ways for enterprises to adopt Generative AI.
Access all research papers, publications, products by Lyzr Research team.
Mixture of Expert Agents Architecture
Lyzr researchers propose a novel ‘Mixture of Expert Agents’ chatbot architecture with ‘hold and pass’ model of handling conversations, enabling specialized and in-context responses to complex customer queries across multiple domains.
Automated Meta Prompting For Better LLM Output
While LLMs like GPT4 and Claude3 are getting better in understanding naive prompts by users, high quality prompts with instructions tend to get better outputs. Lyzr researchers built MagicPrompts module to enable automated improvement of user queries for better information retrieval.
Lyzr Introduces AgentMesh Architecture
As organisations continue to adopt more AI agents, there emerges a need for a data exchange protocol between the agents. Lyzr’s AgentMesh is designed to address this growing concern. AgentMesh allows seamless exchange if AI generated information between agents ensuring the continuous development of ‘General Intelligence’ of the organisation’s AI ecosystem.
AutoRAG for Automated Selection of RAG Pipeline
The retrieval efficiency of a RAG pipeline can be improved with the right combination of various parameters that controls RAG. Lyzr researchers built AutoRAG – an automated way to try 1000+ RAG pipelines for a given data source + test data combination to choose the top RAG pipelines.
