Researchers have developed COGNI, a conversational business intelligence system designed to query diverse enterprise data sources, including structured warehouses and unstructured documents like slide decks. The system employs a four-layer architecture: an indexing layer with slide-adaptive chunking, a routing layer using a fine-tuned Qwen-2.5-1.5B-Instruct model for modality and complexity assessment, a retrieval layer with specialized agents for NL2SQL and multi-hop synthesis queries, and a caching layer for efficient query validation. COGNI achieves high accuracy on internal benchmarks for both data indexing and query routing, while also demonstrating significant cost and latency reductions. AI
IMPACT This system could streamline enterprise data analysis by unifying access to disparate data types, potentially improving efficiency and reducing costs.
RANK_REASON The cluster describes a research paper detailing a new system for enterprise data querying. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
