23 November 2024, Sat |
7:40 AM
Dataiku, a Universal AI Platform company, has introduced its LLM Guard Services suite, aimed at facilitating enterprise Generative AI (GenAI) deployments from proof-of-concept to full production without sacrificing cost, quality, or safety. The new suite includes three integrated solutions: Cost Guard, Safe Guard, and the latest addition, Quality Guard. These components operate within the Dataiku LLM Mesh, recognized as the most comprehensive and agnostic LLM gateway for developing and managing enterprise-grade GenAI applications that adapt over time.
As businesses strive to streamline their operations, a recent Dataiku survey revealed that 88% of enterprise leaders lack specific applications or processes for managing LLMs. Dataiku’s LLM Guard Services addresses this challenge by providing a scalable, no-code framework that promotes transparency, collaboration, and trust in GenAI projects across teams and companies.
Florian Douetteau, CEO of Dataiku, highlighted the industry’s current landscape: “As the AI hype cycle progresses, the initial excitement has transitioned into frustration. The issue lies not with GenAI’s capabilities but its reliability. Ensuring consistent performance regarding cost, quality, and safety is vital for maximizing the technology’s potential in enterprises.” He emphasized that LLM Guard Services facilitates end-to-end management of GenAI rollouts from a centralized platform, helping avoid costly setbacks and the emergence of unauthorized ‘shadow AI.’
Key Features of Dataiku LLM Guard Services:
1. Cost Guard: This dedicated cost-monitoring tool enables enterprises to trace and manage LLM usage effectively, allowing better anticipation of spending versus budget in GenAI initiatives.
2. Safe Guard: This solution evaluates requests and responses to sensitive information, providing customizable tools to secure LLM usage and prevent data abuse and leakage.
3. Quality Guard: The newest component offers automatic, standardized, code-free evaluations of LLMs for various use cases, enhancing response quality while ensuring objectivity and scalability in the evaluation cycle.
Historically, companies deploying GenAI had to rely on custom code-based methods for LLM evaluation or adopt separate, point solutions. Now, with the Dataiku Universal AI Platform, enterprises can effortlessly assess GenAI quality and incorporate this vital step into their GenAI development cycle. By leveraging Quality Guard, customers can automatically compute key LLM evaluation metrics—such as answer relevancy and correctness—using techniques like BERT, Rouge, and Bleu, ensuring they select the most suitable LLM for sustained reliability over time.
