Dataiku LLM Mesh: 90% Reduction in Time to Insights

Title

Dataiku LLM Mesh: 90% Reduction in Time to Insights

Content

Novartis partnered with Dataiku to streamline analytics, machine learning, and AI across the organization, achieving dramatic improvements in time to insights for their GenAI use cases. The collaboration centered on Dataiku's LLM Mesh architecture, which enabled Novartis to deploy and manage multiple large language models through a unified interface. The primary achievement was a 90% reduction in time to insights on GenAI use cases, shrinking the timeline from 21 days to just two days. This was accomplished by implementing a GenAI chatbot powered by Retrieval Augmented Generation (RAG), which used embeddings produced by the Novartis AI pipeline to address user inquiries based on retrieved data and conversational context. Dataiku's platform provided several key capabilities: Prompt Studios enabled iterative prompt refinement for developing high-efficiency prompts; LLM-powered recipes established robust AI pipelines for classification and summarization tasks; and automated metadata extraction streamlined the creation of embeddings for the RAG framework. Beyond the GenAI chatbot, Novartis experienced a 600% acceleration in data ingestion time for spreadsheet-based workflows, reducing processing from one week to six hours. The platform's automation freed up developer resources for more strategic initiatives while improving risk management and governance through transparent AI operations. The implementation democratized data across the organization, empowering both data practitioners and citizen data scientists to build and deploy AI applications without deep technical expertise.

URL